communication and shared understanding in …€¦ · relatively transient. the shared workspace...
TRANSCRIPT
COMMUNICATION AND SHARED UNDERSTANDING INCOLLABORATIVE WRITING
by
Alex Mitchell
A thesis submitted in conformity with the requirementsfor the Degree of Master of Science
Graduate Department of Computer Science
University of Toronto
© Copyright by Alex Mitchell 1996
iii
COMMUNICATION AND SHARED UNDERSTANDING INCOLLABORATIVE WRITING
Master of Science, 1996
Alex Mitchell
Graduate Department of Computer Science
University of Toronto
Abstract
Writing is a form of communication. Writing together is a process of negotiating for
content and meaning of a text. Tools to support collaborative writing must take into
consideration the communication which takes place both around and through the text.
An extended field study of two groups of four grade six students was carried out to
explore the impact of a shared text editor on communication and writing during synchronous
collaborative writing. Analysis of the students' interactions shows that a shared editor affects
control of discourse, the pursuit of attention, and the development of the common ground
essential for shared understanding between collaborating authors.
These observations suggest that effective collaborative writing technology needs to
satisfy design requirements in the areas of concurrency control, awareness, and shared task
spaces. These design requirements were explored through the development of Calliope, a
prototype synchronous collaborative writing tool built using the GroupKit groupware toolkit.
v
Acknowledgments
This thesis emerged from discussion and collaboration with a wide range of people.
Ron Baecker, my supervisor, has been instrumental in keeping me interested in Computer
Science over the past four years. Thanks for encouraging me to push the bounds (without
breaking them), and for pushing me when I needed it...
Janet Salaff, my second reader, introduced me to sociology and helped me write a thesis which
deals with people, not just with computers. After all, without people, what’s the point?
Ilona Posner was my collaborator on the Prejudice Project, and was responsible for most of the
initial work. Her previous work on collaborative writing provided the inspiration for this thesis.
Thanks to the researchers from the CSILE project of the Ontario Institute for Studies in
Education, and the teachers and students at Huron Street Public School, for their cooperation and
enthusiasm. Special thanks to Ryan Fields, Kate Huntly, David Lizoain, Sarah Lyons, Elisa
Morera, Carla Ottens, Faith Tomes and Samantha Yee.
Thanks to Mark Roseman and Saul Greenberg, for GroupKit, and for listening to me complain.
I’d like to thank the reviewers at CHI ‘95 for comments on earlier material. Sara Bly, Hiroshi
Ishii and Karen Holtzblatt also provided valuable insight and feedback.
This research was funded by the Natural Sciences and Engineering Research Council of Canada.
Finally, thanks to my friends and family for putting up with me as I struggled to finish. Yes, it
really is done!
vii
Table of Contents
Abstract iiiAcknowledgments v
Table of Contents viiList of Figures ixList of Tables x
Introduction 1
Part I: Background 51 Related Work 7
Writing Research 9
Collaborative Writing Research 11Perspectives on Communication and Shared Understanding 17Collaborative Writing Systems 21
2 The Prejudice Project 29The Study 29The Participants 32
Setting: The Classrooms 35The Technology 38
3 Data Collection and Analysis 41
Data Collection 41Approach to the Analysis of the Data 43
Part II: Analysis of Writing in a Shared Workspace 494 Patterns of Control Using a Shared Editor 51
Control Away From the Computer 52
Control in the Shared Text Workspace 53Different Control Strategies: Silencing vs. Directing 56Control In and Around the Shared Workspace 58
The Influence of Technology on Control 625 Attention In and Around a Shared Text Workspace 63
The Pursuit of Attention 63
The Computer as an Alternative Focus of Attention 65The Computer as the Primary Focus of Attention 70Attention in Collaborative Writing 76
6 Grounding and Shared Understanding 79
viii
Affordances of a Shared Text Workspace For Communication 80
Costs of Communication Through a Shared Text Workspace 84The Impact of Grounding on Collaborative Writing 87
7 Approaches to Writing in a Shared Workspace 89
A Model of Synchronous Collaborative Writing 89Collaborative Writing and the Models of Individual Writing 93
Part III: The Design of a Collaborative Writing Tool 978 Implications for Design 99
Requirements for a Synchronous Collaborative Editor 99
Calliope: a Prototype Shared Editor 1039 Concurrency Control 109
Approaches to Concurrency Control 110
Concurrency Control in Calliope 11510 Awareness 119
Peripheral Awareness 120
History and Context 12411 Shared Task Spaces 127
Reduce Costs of Grounding 128
Multiple Related Channels of Communication 132
Future Work and Conclusions 137
Appendices 143Appendix A: Notation for Transcripts 143
Appendix B: Samples from Data 145
References 149
ix
List of Figures
1.1: The PREP asynchronous shared editor 221.2: The SASSE collaborative editor 24
2.1: Physical setup of the Prejudice Project 302.2: Floor plan for Room 25, site of the Prejudice Project 362.3: The Aspects shared editor 39
3.1: Data collection setup 417.1: The space of collaborative writing approaches: text and writers 927.2: The space of collaborative writing approaches: text and focus 92
8.1: Impact of shared workspace and resulting design requirements 998.2: Initial version of Calliope 1058.3: Session control using open registration 106
9.1: User-selectable locking levels 1179.2: The main text window in Calliope 11810.1: Telepointer and shared scrollbars 121
10.2: Following a remote user’s view 12210.3: Fisheye overview display 12310.4: Coloured text showing who wrote what 124
10.5: Embedded author information 12511.1: Combining different locking levels 12911.2: Private text 130
11.3: The scratchpad: an unstructured text space 13211.4: Annotation window 13511.5: Attaching a web-based annotation to a point in the text 136
x
List of Tables
1.1: Seven approaches to organization in collaborative writing 141.2: Document control methods 15
1.3: Writing strategies 151.4: Roles taken on during collaborative writing 151.5: Writing activities 16
2.1: Project plan for the first five weeks 302.2: Project plan for the remaining weeks 315.1: The Wednesday group’s focus of visual attention 66
5.2: The Tuesday group’s focus of visual attention 726.1: Affordances for grounding in a shared text workspace 806.2: Costs of grounding in a shared text workspace 84
7.1: Approaches to synchronous collaborative writing 907.2: Writing approaches at different stages in composition of 3 documents 938.1: Summary of design requirements 103
12.1: Comparison of synchronous shared editors based on design requirements 139
1
Introduction
Kate, Carla, Ryan and Sarah are sitting at a row of Macintosh computers,
working in the Aspects1 shared editor on the introduction to their magazine. Thedisplays on their computers all contain the same text document, and anythingtyped on one computer shows up on all the rest. Ryan has typed the first
sentence, and is trying to discuss it with the group."Okay okay we can start it off like something like, okay guys, " Ryan
says, glancing at the rest of the group. He continues, reading off what he has
written so far. "Um. In this magazine you're about to read basically the topics.Um." He stops, again glancing at the rest of the group.
Sarah looks at Ryan and says, "I don't think we should put basically."
Reading from her screen, Carla says, "In this magazine that you are goingto read, we are..."
Looking at his own screen, Ryan interrupts her, saying, "you are about to
read, you..."Without turning from her screen, Carla continues, "we are gonna be
explaining..."
Now looking at Carla, Ryan picks the thread up with, "... about prejudice."Looking straight ahead at her screen, Sarah rephrases the sentence as, "In
this magazine our group has put together the topics are..."
Looking at Sarah, Ryan interjects, "we, very good." He turns to thecomputer, and immediately starts typing.
This brief interaction illustrates three important aspects of collaboration. The group is
trying to compose an introduction to their magazine. As the discussion proceeds, control of the
conversation shifts between the group members, influenced by who is able to make changes to
the text on the computer. The students glance back and forth, and also spend time looking at the
computer screen. The ways in which attention is apportioned between group members, and how
much of that attention is given to the computer, determines the character of the interaction. At
the same time, the emerging sentences need to be grounded in a shared understanding of the
intended content and meaning of the magazine's introduction, the results of the collaboration.
This grounding takes place both in verbal interaction and in the text visible on all the computer
screens.
1Aspects is a trademark of Group Logic, Inc., formerly Group Technologies, Inc.
2
Negotiation for control, the pursuit of attention, and the creation of a common ground for
the group's communication characterize the nature of collaboration within the group.
Writing is a form of communication. To write together, authors need to come to a shared
understanding of the material, and then attempt to reflect this shared understanding in the text.
Collaborative writing forces authors to confront the communicative nature of writing. The
introduction of a collaborative writing tool such as Aspects influences the social context of the
writing task, especially in terms of negotiation for control, attention, and meaning in the text.
Aspects provides a shared workspace on the computers through which authors can compose and
communicate about a text document.
In normal social discourse, communication takes place through both verbal and nonverbal
channels. The information being communicated is put out into the group for negotiation, yet it is
relatively transient. The shared workspace adds a concrete, spatially oriented channel to the
communication, allowing the discourse to be manipulated and viewed in a much different
manner. The focus of the group’s interaction is the shared text workspace provided by the
Aspects collaborative editor. The mechanisms provided by Aspects to permit group members to
control concurrent access to the text in the shared workspace also serve to alter the ways in
which the group members are able to control the discussion.
In this thesis, I will examine how one can look at collaborative writing as a process of
negotiation and communication, see how this relates to existing theories of collaborative writing,
and consider how this approach can be used to inform the design of computer-based tools for
collaborative writing. When people work together on a writing task they can work either
synchronously, with all authors present and working on the task at the same time, or
asynchronously, where the text is worked on by various authors at different times. These
different modes of collaborative writing require much different computer support. I will be
focusing specifically on the interactions which take place during synchronous collaborative
writing.
My discussion will be grounded in a field study of collaborative writing, carried out at the
Huron Street Public School by myself and a co-worker, Ilona Posner, over a period of three
months with two groups of grade six students. The students worked together to write a 32-page
magazine on prejudice, using the Aspects shared text editor. A rich set of data was collected,
including two questionnaires, a variety of artifacts, a final interview and group discussion, and
complete video records of all interactions. The material was approached as an ethnographic
study, with descriptive observations forming the majority of the data.
3
In The Second Self (Turkle, 1984) , Sherry Turkle argues that technology catalyses
change not only in what we do, but also in how we think, and how we construct our images of
self and other. In my study, the computer is an alternative medium of communication, and is
used in the negotiation of text and meaning as the students create a magazine. It also forms a
medium for the negotiation of control by the group. The introduction of the shared space
provided by the computer changes the skills needed for success in a social situation. The fact
that the groups in our study are using both a text-based and a face-to-face medium for
communication makes the social dynamics different than both standard interaction and normal
computer-mediated communication.
The use of a collaborative, networked computer system extends the social dimension to
the electronic world. Although the students are working in a face-to-face situation, they also
make use of the alternative means of communication provided by the computers, with a
corresponding change in social behaviour. The shared workspace, and the nature of the tasks
performed in this study, break down the boundaries of private and public space during creative
writing. The fact that conversation and language, both inherently social processes, are partially
mediated by technology alters the control and power structures in the group. All of these factors
must be taken into consideration in the design of collaborative writing software.
I start by examining the existing literature on collaborative writing, looking at the studies
which have been carried out, the theories developed from these studies, and the tools which these
theories have informed. I then explore how synchronous collaborative writing takes place in a
real situation by performing a field study of students working over an extended period on real
writing tasks. The observations from this study demonstrate the impact of the introduction of
computers on the communication and social interaction which takes place during the process of
writing together. The use of a shared text workspace alters the means used by individuals to
control ideas and discourse in the groups. Rather than just using conversational tools, the group
members are also able to use the control mechanisms provided by the shared editor to control the
discourse.
In addition, the distribution of attention is altered by the presence of the computers. I
examine this by using the tools provided by conversation analysis to look at gaze and attention in
the group interactions. I show that the computers come to act as either an alternative or a
primary focus of attention in the groups, altering the way in which attention is shared between
the group members. This discussion leads to an examination of the ways in which the shared
workspace is used to ground the group’s shared understanding of the content of the developing
text. The presence of the shared text workspace provides an additional channel of
4
communication, and changes the costs and affordances of communication within the group. I
look at these results in terms of the above theoretical background, show how the nature of the
shared workspace and its characterization of communication in the groups affects the way the
groups write together, and present a classification of the writing approaches used in synchronous
collaborative writing.
The results of the study suggest a series of implications for the design of collaborative
writing tools. Systems need to support both communication and task related interactions; I
categorize these requirements in terms of concurrency control, the amount and kind of
awareness of the shared workspace afforded by the system, and the nature of the system’s shared
task spaces. The impact of the computer on negotiation for control of the group's discourse
implies that care needs to be taken in the design of the concurrency control mechanisms provided
by the shared workspace. The role of the shared workspace in the distribution of attention within
the group shows the need to support collaborator awareness, and to provide a seamless interface
between the group's interpersonal space and the shared workspace. Finally, the tendency of the
groups to use the shared workspace as a means of grounding their communication and shared
understanding stresses the importance of exploring design possibilities for shared task spaces
which can be provided by the system.
These design requirements are explored through the implementation of Calliope, a
prototype shared editor. The system is implemented using GroupKit (Roseman, 1996), a
groupware toolkit. I discuss the implementation of the prototype collaborative writing tool,
examine the issues involved, and present some possible directions for future research.
5
Part I: Background
All writing can be seen as inherently collaborative. To motivate my investigation of
synchronous collaborative writing, I begin this section by discussing related work in the study of
individual writing, the observation of collaborative writing, and theories of communication. This
leads to a discussion of the existing computer-based tools designed to support collaborative
writing and the studies of these systems.
Current work in collaborative writing stresses support for the task of writing, at the
expense of the communication which takes place during writing. This motivates my study of
two groups of grade six students working in on extended writing task, the creation of a 32-page
magazine. I present the details of this study, and give some background on the groups and the
technology. Finally, I discuss some of the methodological issues involved in my decision to
carry out an extended, ethnographic study of collaboration to inform the design of collaborative
writing software.
7
Chapter 1
Related WorkAlthough writing is often thought of as an individual endeavour, with writers working in
isolation from the rest of the world, most writing actually involves social interaction. According
to Rubin (1988), all written discourse can be seen as collaborative. There are a number of ways
in which text, like conversation, can be viewed as a collective production. Collaboration can
take the form of written correspondence, which can be explicit, but also includes all texts, which
are in effect utterances in a universe of written discourse arising within the domain of a discourse
community (Rafoth, 1988). Individual writing can also emerge out of conversation, in which
the written word becomes a direct extension of an ongoing dialogue with collaborators. This is
often the case where there is one author, but also a number of people involved in discussing a
topic of common interest. The text can emerge from the discussion, and encourage further
discussion, in an iterative process. Finally, writing can be explicitly collaborative, where a text
is composed through joint authorship. It is this last form of writing that I will be investigating.
Extensive research has been conducted into how people write individually and in groups.
From this research a number of theories of collaborative writing have been developed. This has
in turn inspired the development of computer tools to support the various tasks involved in the
collaborative writing process. To provide some background on the research field, and to set my
research in context, I will now discuss some previous work. I will begin by examining the
approaches traditionally taken by writing researchers. This will be followed by a discussion of
some of the more important studies of collaborative writing and the models of collaborative
writing which have emerged from these studies.
The theoretical background suggests that communication is an important aspect of the
process of writing together. This aspect of group writing has been incorporated into a number of
tools to support groups writing together asynchronously, at different times and places. However,
the importance of communication has not been adequately taken into consideration in the design
of synchronous collaborative writing software. I will look at the existing systems, examine the
research that has been conducted into the way groups use collaborative writing software, and
suggest the ways in which these systems and studies are inadequate. This will in turn motivate
my approach to the study of collaborative writing and the design of computer-based tools to
support people writing together.
8
Writing ResearchWhen looking at collaborative writing it is useful to examine the insights drawn from
traditional writing research. Early studies of individual writing focused largely on product. The
standard attitude towards the writing process characterized it as an “outline-draft-edit” cycle,
with revision limited to word-tinkering. This assumes that meaning is clarified in the head first,
and then articulated clearly to express that meaning in text. There was a shift to process in the
late 1970s (Freedman, Dyson, Flower and Chafe, 1987) with the development of cognitive
process models of composing. Writing was thought of as a process of problem solving involving
a sequence of activities, including setting goals, planning, organizing, transcribing and editing
(Flower, Schriver, Carey, Haas and Hayes, 1989; Hayes and Flower, 1980). It is this research
from which collaborative writing research tends to draw most heavily. In the 1980s the focus of
writing research shifted again, this time placing the emphasis on the context of writing tasks,
with meaning coming to be seen as an ongoing negotiation between the writer and the readerwith the goal of attaining a shared socially constructed reality (Nystrand, 1989) .
Fitzgerald (1992) presents a comprehensive overview of the different approaches to the
study of individual writing. She identifies three models of writing: the stages model (Rohman,
1965) , the problem-solving model (Flower and Hayes, 1984) , and the social interaction model
(Nystrand, 1989) . She sees these models as distinguished from each other not only by their
characterizations of the writing process, but also by their underlying views of knowledge and
knowing. She claims that the assumptions inherent in the researcher’s attitudes towards
knowledge have an inextricable impact on the theories developed.
Knowledge and Writing
Fitzgerald feels that beliefs about knowledge and ways of knowing construct and direct
attention to certain problems and methodologies, but asserts that the different beliefs are not
incompatible, and should be used to complement each other. According to Fitzgerald, there are
two major conceptions of what is meant by knowledge: the exogenic and the endogenic. The
ways in which these paradigms view knowledge can be split into spacial and temporal
dimensions, focusing on what is meant by “knowledge” and what is meant by “knowing”.
The exogenic viewpoint is often associated with logical positivism, realism, experimental
research with generalization, and the works of philosophers such as Bacon, Locke, Hume,
Hobbes and Mill. This paradigm sees knowledge as located in the world, and mirrored in the
mind. Knowledge is seen as static and unchanging, existing independent of the “knower” in
terms of facts and a discoverable “Truth”. There is little or no link between knowledge and the
process of knowing. Knowledge is discovered, with knowing an objective process and the
9
knower at a distance from the facts.
The endogenic point of view can be linked to ethnography, phenomenology,
hermeneutics, symbolic interactionism, constructivism, and the works of, for example, Descartes,
Leibnitz, and Kant. The endogenic paradigm sees knowledge as located in the mind, changeable
and not independent of the knower. Knowledge is still seen as facts, but also as cognitions,
feelings or emotions, and there are possibly various “correct” versions of the truth. Knowledge
can be created or made, and is inherently subjective.
Fitzgerald also identifies a third, emerging paradigm, which could be seen as social
constructivism or a combination of constructivism and symbolic interactionism. Symbolic
interactionism sees the world as generated by social interaction which itself produces and is
shaped by participants’ interpretations. Constructivism sees knowledge as constructed by
individuals through direct interaction with the world. The third point of view is similar to the
endogenic in the view that knowledge is constructed, and is more than just facts. The difference
is that this paradigm doesn’t have the dualistic quality of the first two paradigms. Knowledge is
neither objective nor subjective; rather it consists of the interaction of the subject and object.
The “object” is no longer a given, but is a product of this interaction. In the process of knowing,
the observer, the observed, and the process of observation are a totality, and cannot be separated.
Models of Individual Writing
With these different views of knowledge and knowing in mind, Fitzgerald asks how the
researcher’s view of knowledge affects her approach to writing research. The world in which the
writing process takes place can be divided into three regions: the writer, the reader, and the text.
These regions are surrounded by the knowledge, skills, feelings and cognitions on which they
draw. The three models of writing have different views of the interaction between these regions,
views which influences the problems addressed and methodologies applied in research.
The Stages Model
The exogenic stages model is highly linear, with the various writing stages seen as
separated in time. The stages involved are pre-writing (idea generation), writing (text
generation) and rewriting (edit/rework text). The focus is on the text as an external object.
Writing is seen as knowledge transmission, a process of finding and structuring information and
translating this information into words independent of the writer’s ideas.
The Problem-Solving Model
The problem-solving model (Flower and Hayes, 1984) lies between the exogenic and the
endogenic, and is still closely positivistic in terms of methodology. The model attempts to depict
10
the mental processes involved in the writing process. Writing is seen as a process of problem-
solving, with a set of goals and purposes to be achieved by the writer. The writing processes
consist of planning, translating, reviewing and monitoring the developing text. When the text
and goals are at odds, methods are used to solve these discrepancies. This forms an informal
information-processing model of writing. The task environment is the problem and the text so
far. Long term memory holds the writer’s knowledge of the topic, the audience and the plans for
the text. Writing takes place according to the following procedure:
1. writers identify discrepancies between the intended and instantiated text, i.e.
they have goals and monitor for inconsistencies2. writers diagnose problems and determine changes needed3. writers operate to carry out required changes
The problem-solving model sees the text as constructed in the mind; however,
researchers tend to take an inherently external view of the process, often using methods such as
think-aloud protocols to investigate the internal processes. The concern is with the product and
process, and is highly rule-based. The author is placed at a distance from the goal, and there is
no reader involved.
The Social-Interaction Model
The social-interaction model of writing (Nystrand, 1989) characterizes writing as a
negotiation between the writer and the reader for meaning in the text, and is closely related to
social constructivism. Writing is seen as a mental and constructive process, an interaction of the
minds involved rather than just an internal process in the individual mind of the author. Written
communication is a “fiduciary act for both writers and readers in which they continuously seek
to orient themselves to a projected state of convergence between them.” (p. 75) This model is
close to the new, third paradigm of knowledge and knowing, but Fitzgerald argues that it does
not totally fit, since it still sees the object as a given, rather than the result of social interaction.
According to the social-interaction model of writing, composition involves three iterative
phases. Writers first initiate the written discourse; writers try to fashion the beginning of a text
to establish a mutual frame of reference or shared social reality between the reader and the
writer. The writer will introduce new information, and test for reciprocity. If the new
information threatens reciprocity, it is a trouble source, and presents a choice point for text
change. Writers then attempt to sustain the written discourse. Failure to fix the trouble leads to
misconstructs. Writers employ options or elaborations to clarify the communication in an
attempt to restore the shared social reality.
Problems arise as incongruities between readers and writers, and “solutions” are
11
elaborations, “tentative text hypotheses which the writer puts forth as a sort of test for the
reader's response” (Nystrand 1986, p. 204) . The problem is a gap of understanding to be
bridged, and the goal is a shared social reality, rather than the text itself. Writers and readers
work together to attain shared understanding. The problem is defined socially, not absolutely,
and knowledge is a socially constructed artifact in a community of writers and readers.
Knowledge is located in the minds of communicating partners, not in an isolated author.
“Communication occurs when a writer's elaborations mesh with a reader's expectations.”
(Nystrand and Brandt, 1989, p. 219) Knowledge depends on the minds involved, and writing is
a situation-dependent constructive process.
Implications for Collaborative Writing Research
Although appearing to be in conflict, these models can in fact be seen as complementary,
providing different perspectives on the same problem. The stages model focuses on the
construction of the text itself. The problem-solving model steps back and looks at the writer in
context. Finally, the social-interaction model makes linkages between the writer and readers,
and addresses the social aspects of writing.
The social-interaction model of writing sees writing as a process of interaction between
the minds involved in the writer-reader-text interaction. The writer is no longer seen as an
isolated individual, and the process of writing is no longer an abstract, internal process. Rather,
it is a situated social process which affects and is affected by all the participants. This
characterization of the writing process suggests an interesting way to look at collaborative
writing; when there are several authors involved, the social nature of writing is made explicit,
and communication and negotiation for shared understanding become even more crucial. It is
the social interactions between writers and writers as readers, as well as the mechanics of the
task, which must be examined and accounted for in the design of computer support for
collaborative writing. The importance of the social aspects of writing will become clear in the
following sections.
Collaborative Writing ResearchA fair amount of observational research has been conducted into the way groups write
together. These studies have generally involved interviews with or surveys of groups of
experienced writers who have worked on several projects making use of conventional writing
tools. The concern has been with the organization and coordination that takes place during group
writing, and the division of roles and responsibilities within the group. These studies provide
insight into collaborative writing, suggesting that there is a need for flexibility and
communication throughout the entire collaborative writing project. There have been several
12
attempts to develop theories that describe how people write together, focusing on the
organizational structure, roles, and strategies of groups working on a joint writing task. These
theories show the importance of a variety of interpersonal factors which have an impact on the
process of writing together. It is not just the performance of the task, but also the interaction of
the participants, which needs to be addressed in the support of collaborative writing.
Observational Research
An important part of the process of writing together is planning and coordination. Based
on a series of 20 interviews covering a total of 14 different collaborative writing projects, Allen,
Atkinson, Morgan, Moore and Snow (1987) found that all the groups interviewed conducted
face-to-face planning sessions at the start of a project. Further into the project, different writing
approaches were used by the various groups. Kraut, Egido and Galegher (1990) performed an
interview study, a survey, and an archival study of scientists who co-authored journal articles.
Results suggested that the participants felt that the establishment and support of group
relationships was at least as important as the content of the work itself. The general pattern, as in
the Allen et. al. studies, was for collaborative projects to be jointly planned and individually
executed. Both physical proximity and informal communication were important for the
initiation, maintenance, and completion of joint projects.
In an extensive study, Ede and Lunsford (1990) carried out both a questionnaire survey
and interviews with 700 professionals in seven fields, 87% of whom were discovered to write
jointly some of the time. 58% of these found joint writing productive. Again, communication
and coordination were seen as having a major impact on the writing process, with the large
amount of time required for coordinating the writing cited as one major drawback to
collaborative writing. The organizational strategy chosen was determined to be closely related to
successful results. Ede and Lunsford identified two approaches to organizing collaborative
writing: hierarchical collaboration involves a highly structured organization, with goals and
roles well-defined, whereas dialogic collaboration is more loosely structured, with fluid goals
and multiple, shifting roles.
These studies provide a broad view of how people write together on extended joint
projects. Planning and coordination are important, and the choice of organization and strategy in
the approach to these factors affects the success of the writing project. Posner (1991; 1992)
focuses closely on the joint writing process, looking at how the text is created and controlled in
the context of the events leading to the completion of the writing project. From a series of
interviews with 10 individuals who had participated in 22 collaborative writing projects, Posner
observed that groups tended to move between different approaches as needed, depending on the
13
structure of the group and the context of the task. The lack of support for these transitions and
the various stages of writing were seen as deficiencies in existing technology. Again,
coordination and communication between group members across the course of the project was an
important consideration.
Beck (1993) conducted a survey of the experiences of 23 largely academic co-authors.
Flexibility was found to be important, as participants constantly renegotiated roles, content and
responsibility. Beck found that participants took part in discussions more during the writing
process than they did before or after the actual writing of the document. Beck came to
characterize collaborative writing as a “dynamic process with continuous negotiation and re-
negotiation of questions relating to both the contents of the document, and ... leadership roles and
sharing of responsibilities between the co-authors.” (p. 87) This suggests that the
communication that takes place during group writing involves not just the task itself, but also the
social context of the task of writing together. This is supported by Kaye’s (1993) examination
of three case studies of group authoring, which stressed the impact of social, psychological and
institutional factors on the success of collaborative writing endeavours. Dillon (1993) adds the
observation that computer support should focus on providing communications support for the
sharing of documents and contacting collaborators.
Beck and Bellotti (1993) further explore the need for flexibility in the support for
collaborative writing. Their study looks at three in-depth case studies of academic collaboration,
and emphasizes the need for communication about the content of the developing text and the
context of the writing task. Roles constantly shifted throughout the writing process; writers
would switch from working closely to segmenting the document and working separately as was
appropriate and convenient. This required the authors to coordinate both the editing of the text
and the responsibility of the co-authors. Document structure helped provide a point of reference
for this coordination. Editing coordination relied on the ability to visualize the document,
ground communication in the content of the document, communicate changes, and show
authorship and document history. Coordination of responsibility necessitated planning,
awareness of the status of text and ongoing activities, negotiation of access to the text, and
flexibility in terms of roles and responsibility. Beck and Bellotti suggest that systems should
help inform, not constrain, collaborators as they negotiate the development of a jointly authored
text.
Theories of Collaborative Writing
These studies have informed a number of theories of collaborative writing. Based on
their extensive survey of authors who worked on joint writing projects, Ede and Lunsford (1990)
14
developed a categorization of the organizational patterns used by collaborative writers (see Table
1.1). This categorization captures the different approaches taken by groups to the issues of
planning, organization and division of the writing task, responsibility for the document, and
revision of the final product. Ede and Lunsford closely follow the stages model of writing,
looking at how the group plans, outlines, writes the draft, and revises the document.
Pre-writing Writing Rewriting
The group plans and outlines. Each member drafts a part. Group combines the parts andrevises the whole document.
The group plans and outlines. One member drafts. The whole group revises.
One member plans and writes draft. The group revises.
One person plans and writes draft. One or more persons revise(s)
the draft without consultingthe first writer.
The group plans and writes draft. One or more person(s) revises
without consulting the writers.
One member assigns writingtasks.
Each member performsindividual task.
One person combines thesesegments and revises the
whole document.
One person dictates. Another person transcribes thedictation and revises the text.
Table 1.1: Seven approaches to organization in collaborative writing (Ede and Lunsford, 1990) .
This description of collaborative writing, while applying to hierarchical collaboration,
does not reflect the flexibility that is needed for the more fluid and constantly changing roles and
strategies used during dialogic collaboration. Both Posner (1991; 1992) and Sharples, Goodlet,
Beck, Wood, Easterbrook and Plowman (1993) have developed taxonomies of collaborative
writing to describe the various ways in which groups write together. Posner provides four
different perspectives for approaching the way in which a collaborative writing task is
performed: roles, activities, document control methods, and writing strategies . A similar
description of collaborative writing was presented by Sharples et. al., where issues important to
collaborative writing are divided into coordination strategies and collaborators’ roles.
Sharples et. al. categorize the coordination strategies used by joint writers into parallel,
sequential and reciprocal approaches to joint work. These can be interpreted in terms of
Posner’s document control methods (see Table 1.2) and writing strategies (see Table 1.3).
Parallel work involves the division of writing into subtasks to be accomplished in parallel.
15
Collaborators work simultaneously, and send their products to each other, with is no central
coordinator. Posner would describe this as using a separate writers approach to writing with
independent document control. Sequential work involves collaborators dividing the task so that
the output from one stage is passed to the next writer in line. Each stage may be a section of the
text or a complete draft. This can be seen as separate writers with relay document control.
Reciprocal work involves the joint creation of a common product. A complete draft of the text is
produced either by composing together out loud with one person writing down the contributions,
or by all collaborators putting ideas freely into a central text. Posner describes this approach in
terms of writing strategy, which could be either scribe or joint writing, and in terms of document
control method, which could be either centralized or shared.
Centralized one person maintains the document while
others make suggestions to the writer
Relay one person controls the document at a time,but control passes between multiple authors
Independent several people support segments of thedocument, while each one maintains controlover an individual segment
Shared several people jointly control the document,all having equal access and write privileges
Table 1.2: Document control methods (Posner, 1991)
Single writer text of document reflects thoughts and style of
one individual with minimal assistance fromothers
Scribe most often used in group meetings when one
individual records the group’s discussionswith minimal guidance from the group
Separate writers team members take different parts of the
document and write them individually; laterthe parts are combined to form a whole
Joint writing several team members compose the text
together, where each word choice andsentence structure is decided through a groupeffort
Table 1.3: Writing strategies (Posner, 1991)
Roles include facilitator, scribe, consultant, reviewer, and editor (see Table 1.4). This
16
classification suggests a rigid division of roles between group members. This is sometimes the
case; however, as Posner cautions, even in a long-term, formal collaboration roles tend to change
over time. Participants can switch roles throughout the process, and writing approaches and
document control methods are often changed to suit the needs of the writing task.
Facilitator organizes the division of tasks and assignment
of roles but does not write the text
Scribe converts ideas into text, records the text, freelymakes changes to the text
Consultant actively participates in different stages of theproject but does not write the text
Editor makes corrections to the text written by
someone else
Reviewer provides comments about the text of thedocument
Table 1.4: Roles taken on during collaborative writing (Posner, 1991)
Posner uses a classification of writing activities that closely resembles the stages model of
writing (see Table 1.5). All of these activities can take place at any time in the process of
creating a text. The roles taken on by collaborators and the activities performed are closely
interrelated.
Brainstorming generating ideas
Researching gathering information from sources external tothe group
Planning creating an outline for the document, and oftendividing the work among group members
Writing transforming ideas into text
Editing making changes to the written text
Reviewing generating comments about the text
Table 1.5: Writing activities (Posner, 1991)
These taxonomies describe the organization and coordination that takes place during
collaborative writing. Sharples’ taxonomy, while useful, does not provide as much descriptive
power as the one developed by Posner. In both taxonomies, however, other than in terms of
control of the document, there is no mention of the written representation of ideas essential to
writing, nor is there any mention of the nature of communication between the group members.
Sharples and Pemberton (1992) discuss the means by which a writer interacts with other
17
media such as books, notes, pencils and paper, and the ways in which these media affect the
process of idea generation and written composition. Sharples and Pemberton argue that the
external representation of ideas related to the text being composed constrain the process of
writing, and suggest that a variety of tools should be provided to allow writers freedom to choose
appropriate representations. They distinguish between instantiated representations, which
involve the presentation of ideas in connected prose; uninstantiated representations, idea-labels
that represent an place-holder for as-yet unformed text; and annotational representations, which
represent comments expressing intentions or assessments of existing text. They also examine the
views which are created of developing texts. Items can be unorganized, as in a brainstorming
session; follow a non-linear organization, such as a cluster of labels; or form a linear
organization, such as an outline or table of contents. Sharples et. al. (1993) suggest
collaborating authors are more likely to externalize the intermediate states of a document, and
communicate these to co-authors. Issues include what to represent, the constraints placed on the
document by the representation, the means of communicating changes and comments, and the
effects of media on external representations.
Newman and Newman (1992) take a different approach to the task of writing, looking at
the type of discourse entered into by the text, and the effects of the type of content on the writing
process. The authors identify 3 modes of collaborative writing: literature, documentation, and
critical discourse. Depending on the type of text being written, there is a different relationship
between the text and the negotiated construction of reality. In literature, the writers negotiate the
common definition of reality, reach closure, and then write. With documentation, the definition
of reality is already set, and the writers often have little role in its negotiation, so the writers just
divide up the task, write the text, and then bring the material back together. In critical discourse,
as with literature, there is a negotiation of meaning, but writers can also revise this shared reality
during the editing process, with the text often forming a kind of “metatext”. There is a different
focus of negotiation in the different modes. Problems of group coordination and the negotiation
of consensus arise in all, but to different degrees. With documentation the focus is on accuracy
and conformity to conventions of form and style, whereas with critical discourse the focus is
more likely on ill-defined issues such as competence and ontology, with a much greater
possibility of misunderstandings and conflict.
Newman and Newman argue that the design of collaborative writing systems, which
often focuses on the basic level of resource sharing and tools to manipulate the text, should also
provide higher level support for the negotiation of ideas and construction of shared social reality.
Specifically, the theoretical framework and tools developed to support collaborative writing
18
don’t take into consideration the communicative nature of collaborative writing, the negotiation
for shared understanding and the constant revision and shifting of meaning that takes place
during writing until a consensus is reached about and within a text.
Perspectives on Communication and Shared UnderstandingThe above material suggests that communication is an important part of the process of
creating meaning within a text, especially when there are several authors involved in the creation
of the text. Written communication is different from speech in several fundamental ways
(Nystrand, 1982) . This is mainly due to the context and character of the composing process.
Writing is generally a private, solitary activity, the purpose of which is to communicate with
someone who is not immediately present; speech, on the other hand, is a public activity
involving continuous give and take between conversants who can support the speaker’s
utterances by showing understanding, or show lack of understanding by asking for clarification.
Writing involves no explicit feedback, whereas speaking involves continuous feedback.
Scardamalia, Bereiter and Goelman (1982) focus on this difference as a way to
characterize the ways in which writing differs from speech. Writing is less time-constrained,
with no immediate social situation which, in speech, requires constant monitoring. With writing,
there is no need to coordinate the production of utterances with gestures and eye contact,
whereas in speech there is both the necessity and the ability to make use of feedback from the
conversational partner. Cooper (1982) argues that, in writing, the means of communication is
more concrete and long-lasting than in speaking, so writing can be considered a less situated
communicative act than speaking. However, in a face-to-face collaborative writing task,
communication takes place about the immediate task of writing a text. This makes writing
together more of a situated communicative act than individual writing. To gain further insight
into the communicative nature of collaborative writing, I will now examine some of the
theoretical basis for the study of human communication.
Communication and Meaning
The study of communication is the study of how meaning is conveyed (Leeds-Hurwitz,
1993) . The world is a rich and complex place, full of subtleties of meaning. People are
responsible for creating the meaning in the world, through minor details, both verbal and
nonverbal. Meanings change as they are used, and as required. There is a close relation between
signs (which make up codes) and codes (which form cultures). An important part of interaction
is nonverbal communication, including facial expressions, eye gaze, the space between
participants, and touch. Culture emerges through interaction, revealing the connection between
the individual and the social, the private and the public.
19
Communication is “the symbolic process whereby reality is produced, maintained,
repaired, and transformed.” (Carey, 1975, p. 10) Signs are the smallest element of meaning in
interaction. Codes are sets of related signs and the associated rules for their use.
Communication uses symbols, which are particular instances of signs, and involves the
combination of symbols into codes. Those codes are used to socially construct reality, which
permits ordered human interaction. To study communication, one has to study “texts” to
understand how the larger entity, culture, operates. In this context, a text is a discourse, which
can consist of anything from a single sentence to an extended story cycle.
People live in a world of symbols, not things. It is these signs and symbols which give
the world meaning. Signs are used to create a social reality, the overlay of meaning laid across
the natural world. The sign can be seen as either a dichotomy, consisting of signifier and
signified, or a trichotomy, composed of sign, object, and interpretation. It is this latter, symbolic
interactionist view which I will be using. From this perspective there is a close link between
interpretation and meaning: a text does not necessarily exist independent of interpretation. This
introduces the problem of polysemy, or multiple meanings for one symbol. Interpretation comes
from those who see a symbol, not the symbol itself. Meaning is seen in relation to the
interpretation and the sign itself. Ultimately, meaning resides in the sign, the context of the sign,
and the social actions surrounding the interpretation. Codes obtain meaning from the
relationships between signs, and are combined to form a message or text. Gaps and
inconsistencies are part of the nature of codes. Meaning comes from agreement and shared
cultural context and experience between the participants in a discourse.
Any communication must be based on actual knowledge, beliefs and a set of assumptions
within which the conversation takes place; grounding is the process of seeking and providing a
common understanding around which the conversation takes place. In the message model of
communication, it is assumed that all participants, including passive listeners, share a common
ground (Reddy, 1979) . This does not adequately describe the process of developing a common
ground. The contribution model introduces the idea of presentation and acceptance phases of
conversation (Clark and Schaefer, 1989) . As a conversation takes place, participants offer ideas,
and make repairs or offer alternate descriptions in response to indications of lack of
understanding. The development of common ground takes place as a continuous process,
constantly correcting and revising the shared social reality between the participants. This shared
understanding must be internalized and somehow represented as knowledge in the minds of the
participants (Resnick, 1991) .
20
Knowledge in the World
However, as suggested by Vygotsky (1978) , knowledge is not represented solely in the
mind, but is also represented in the world around us. Cognitive processes do not occur only in
the mind, but also in artifacts in the world. Knowledge and understanding actually take place in
the world. Meaning and understanding are negotiated in and around artifacts in the world, and
are created through social interaction.
Flor and Hutchins (1991) explore the notion that knowledge is shared between
individuals and the artifacts that form the environment in which collaboration takes place. These
artifacts form external structures or schema which are transformed when operated on by
individuals. The collection of individuals and artifacts that participate in the performance of a
task can be viewed as a complex cognitive system. This system can be used as an external
structure which comprises its “mental state”. The collaborative task can be viewed as a complex
cognitive task distinct from the individuals involved. Behaviour results from the interaction
between external and internal representational structures.
From the distributed cognition point of view (Zhang, 1994), the complex cognitive
system performs the task, not the individual; since the system consists of external
representations, it is available for inspection. These representations reflect the need for shared
knowledge and common ground between the participants in a collaborative task. Actors make
assumptions about each other, and about the knowledge and behaviour of the system artifacts.
When shared expectations are violated, corrections and reinforcements must be offered. The
process of negotiating plans distributes them between the actors; there is a sharing of goals and
plans, as a greater common ground is developed. These ideas are related to the theory of situated
cognition (Suchman, 1987) , which suggests that the activities of individuals are situated in the
social and physical contexts around them, and that knowledge can be considered a relation
between the individual and the situation.
Communication is the process of the construction of meaning in society. The structure,
not just the content, of a communication has an impact on the message and the meaning.
McLuhan (Neill, 1993) has suggested that media, and particularly technological media, affect
the ways in which people communicate. Seeing artifacts as part of language, the way in which
objects are used alters the ways in which people communicate. McLuhan views communication
as a process of transformation; the use of an object, or medium of communication, is part of the
experience. In conjunction with medium, the content, and the viewer a meaning is formed.
Grounding in conversation takes place in within the medium of communication used by the
conversational participants. Clark and Brennan (1991) suggest that participants tend to use the
21
medium that requires the least effort. They characterize different media as having a number of
constraints and costs which affect interaction and communication.
Csikszentmihalyi (1981) explores the socializing effect of artifacts. Objects in a
person’s environment tend to create a commonality, helping to create, express and communicate
personal qualities. Inanimate objects can be considered an “other” in a social situation, playing a
role in defining the identity of people who interact within that situation. Objects affect what a
person can do, expanding or restricting the person’s scope of actions and thoughts. Objects help
to create shared meanings; there is a development of a common sense of objects, a community
around objects. Objects convey information about the object’s owners, surroundings, and
context. Artifacts can be used to control and shape the world, and to get in touch with it. This is
especially true for technology.
Turkle (1984) argues that technology catalyses change, not only in what we do but in
how we think. Computers can be seen as an expressive medium. The computer plays a role in
the construction of our physical and psychological world. From the symbolic interactionism
perspective, the computer has a place in the interaction and negotiation that creates the world
around us. Computers as artifacts have become a part of our culture and society, playing a role
in the construction of self, as people come to define themselves as different from the computer.
Turkle sees the computer as a kind of Rorschach; it allows the expression of personality, and is
both a projective and a constructive medium. The computer acts as a marginal object, not
separate from our expression of self, evoking unconscious memories of the zone between self
and nonself. As a new expressive medium, the computer is a challenge, a mirror in which to
reflect ourselves and with which to create shared meaning in society.
Collaborative Writing SystemsThe above discussion suggests that the introduction of computer tools to support
collaborative writing will have a significant impact on the communication which takes place
during the process of writing together. Collaborative writing involves several people working
together, sometimes in the same place and sometimes at a distance, sometimes at the same time
and sometimes at different times. Systems designed to support collaborative writing have been
traditionally divided into two categories, based on whether writing by different authors is
synchronous, taking place at the same time, or asynchronous, happening at different times (Ellis,
Gibbs and Rein, 1991) . This division is not absolute, with some semi-synchronous systems
falling in between. In my thesis I am focusing on synchronous collaborative writing in a co-
located situation; however, since collaboration is never strictly synchronous or asynchronous, I
will include a brief mention of semi-synchronous and asynchronous systems to illustrate the
22
range of systems that have been developed, and the approaches typically taken by designers.
Asynchronous systems
Asynchronous systems tend to support groups that work in a sequential manner, passing
versions of the document around between group members. This requires support for
communication of comments and changes within the document. Quilt (Leland, Fish and Kraut,
1988) is intended to support communication and information sharing among collaborators.
Quilt provides a variety of text and voice annotations, and uses a complex model for managing
and controlling the writing process. This structured model of communication allows suggested
revisions, public comments, directed messages, and private comments to be made by co-authors,
commentors, and readers. Roles are defined explicitly within the system, and determine each
collaborator’s create, modify, delete, annotate, and read permissions. Communication is
included as an underlying part of the design of the system, and is afforded both through the
document and by interaction with external channels of communication such as electronic mail.
Another system designed to support asynchronous collaborative writing is PREP
(Neuwirth and Kaufer, 1989; Neuwirth, Kaufer, Chandhok and Morris, 1990) , which can be
viewed as a “spreadsheet for documents” in which successive paragraphs of a text are arrayed in
adjacent cells vertically, and in which successive versions of the text, or comments on the text,
are arrayed in adjacent cells horizontally (see Figure 1.1). Columns can be used for versions of
the text, plans for the text, and comments or annotations on the text. As with Quilt, support for
communication was seen as an important part of the design of PREP. However, unlike, and in
fact in reaction to, the approach taken by Quilt, roles are not defined explicitly within the system;
rather, the roles of the participants are defined on a social level. This reflects the need of groups
to constantly redefine and renegotiate the roles and responsibilities of group members during the
course of a collaborative project. PREP also attempts to deal with the external representation of
text, placing comments in a distinct but equal position with respect to the main text.
23
Figure 1.1: The PREP asynchronous shared editor.The figure shows a content column (left), and a comment column (right).
More recently, PREP has been extended to incorporate a variety of methods for
displaying the differences between successive versions of a document (Neuwirth, Chandhok,
Kaufer, Erion, Morris and Miller, 1992) . PREP also allows asynchronous communication
between authors through a document transmission facility, thereby affording collaboration
between distributed co-authors (Neuwirth, Kaufer, Chandhok and Morris, 1994) . This
communication attempts to support the wide range of interaction patterns of collaborative writing
groups and accommodate the changing writing strategies, roles and communication needs across
the course of a group writing project.
Semi-synchronous systems
In practice, there is no distinct split between synchronous and asynchronous work. A
number of systems have been developed that attempt to address this fact. CES (Greif, Seliger
and Weihl, 1992) was a semi-synchronous collaborative editor in which authors were able to
work together in different sections of a single document. Changes were propagated at a delay to
all copies of the document using traditional concurrency control techniques. Authors would
gradually be made aware of changes made by remote authors. A trade-off was made between
responsiveness of the interface and notification of remote changes. This is not a problem when
authors are working in different locations and do not need to be made aware of others’ activities
immediately, but can lead to problems when authors attempt to discuss content and use the
shared workspace as a means of grounding the discussion.
Minör and Magnuson (1993) describe a system which supports a mix of synchronous
and asynchronous collaborative writing, in an attempt to bridge the gap between these two
24
modes of collaborative writing. This is done by hierarchically partitioning the document and
using fine-grained version control. The aim is to provide different levels of concurrency to suit
different writing strategies, with the ability to move seamlessly between different modes of
writing. As with CES, the structure of the document is an integral part of the way in which the
text is shared between authors, determining the degree of concurrent access and notification of
changes.
Duplex (Pacull, Sandoz and Schiper, 1994) is an attempt to support varying degrees of
asynchronous interaction in an manner which allows for dynamic decomposition of the
document, rather than relying on the hierarchical structure of the document. Control of
concurrent access is adaptive, changing to account for the way in which the group is working.
This allows for flexibility in terms of working styles, but does not allow for truly synchronous
work, which can be essential during brainstorming and editing.
Synchronous systems
Interestingly, there have been quite a few systems developed to support synchronous
collaborative writing. This is interesting because the form of joint writing supported by such
systems does not really match any of the forms of writing possible using conventional
technology. The ability to concurrently view and access the same text in several different
locations has the potential to significantly alter both the character of the task of writing together
and the communication that takes place between co-authors.
GROVE (Ellis, Gibbs and Rein, 1989; Ellis, Gibbs and Rein, 1991) was a real-time
outline editor allowing simultaneous editing of private, public, and shared views of a document
outline. GROVE attempted to support a shared context, in the form of a common text space,
with individual views of the shared document. A complex algorithm provided serialization of
changes, allowing users to edit all text simultaneously with no locking and no loss of
consistency. GROVE also allowed users to explicitly change access to the text at different levels
of the outline hierarchy.
ShrEdit (McGuffin and Olson, 1992; Olson, Olson, Mack, and Wellner, 1990) supports
the real-time concurrent editing of documents on a number of Macintoshes on a local area
network. Conflicts are prevented by selection-level locking of portions of the document that are
being modified by others. Coordination and collaborator awareness are enabled through the use
of a tracking and control window, which enables users to request information about where other
users are and what they are doing.
25
SASSE (Baecker, Nastos, Posner and Mawby, 1993; Baecker, Glass, Mitchell and
Posner, 1994) is a Macintosh-based prototype shared editor that supports synchronous outlining,
writing, and editing of texts over local- and wide-area networks (see Figure 1.2). SASSE
attempts to enhance the basic shared object with both a variety of mechanisms to support
awareness of the participants in a group project, and a variety of ways of viewing and
manipulating the shared object. Collaboration awareness is enhanced through a variety of
mechanisms utilizing colour, split screen display, non-speech audio, an observation view which
allows monitoring of the work of a collaborator, and a gestalt view which provides an overview
of the document. Both tightly coupled work, via a What You See Is What I See (WYSIWIS)
mode, and loosely coupled work, via unlinked views, are supported. A check in and check out
facility and an interface to the CSILE collaborative knowledge building environment
(Scardamalia and Bereiter, 1992) provide the capability for asynchronous work.
Figure 1.2: The SASSE collaborative editor.The figure shows the gestalt view, colour-coded remote selections and shared scrollbars.
For my study, I intended to use SASSE. However, due to the fact that SASSE was an
untested prototype system, this proved to be impossible. Instead, I chose to use Aspects (Group
Technologies, 1990) , a real-time writing, drawing, and painting system that links multiple
Macintoshes over local- and wide-area networks. Three mediation levels are supported: one
26
writer, in which others can only watch; one-at-a-time, with a turn-taking mechanisms; and free-
for-all, which provides concurrent access to the document under a paragraph-level locking
scheme. I will provide more details on Aspects in the next chapter.
Studies of the Usage of Collaborative Writing Systems
A number of previous studies have been carried out to observe the use of computers to
support collaborative writing, most extensively with the ShrEdit system. However, none of these
studies provide insight into the use of the tools in real world, extended projects. Instead, they
focus on the use of systems in tightly controlled experimental situations.
The ShrEdit studies looked at the use of a shared editor in problem solving and
brainstorming, and focused on trying to show that using a collaborative writing tool improves
performance. Olson, Olson, Storrøsten and Carter (1992) reported how the use of ShrEdit
changed the way groups of people carried out a problem solving task, the design of an automatic
post office. Groups using pencil and paper or a whiteboard were compared to those who had
access to these tools but were encouraged to use ShrEdit. Although the impact of the
collaborative writing tool was confounded with the impact of a computerized writing tool, the
study produced interesting results such as a significant 20% improvement in document quality
for the technology-supported group. Detailed analysis of the typing transcripts revealed that
simultaneous typing occurred, on average, 10% of the time. This is interesting, but does not tell
us much about how the presence of the collaborative writing system influenced the way the
groups communicated or wrote together. The focus was largely on the resulting product, and on
statistical measures that have little relation to the process of creating meaning in a shared text.
McLaughlin Hymes and Olson (1992) studied the use of ShrEdit as a vehicle to enhance
group brainstorming. Although research typically finds that brainstorming groups do not
perform as well as groups working individually and then pooling their ideas, the use of ShrEdit
allowed parallel interacting groups to perform as well as nominal groups. Olson, Olson and
Meader (1995) continued this research to examine the differences between face-to-face groups,
remote groups using an audio connection and remote groups using audio and video. The results
suggested that video supported remote groups can perform at the same level of ability as face-to-
face, but that audio only groups suffered, possibly due to perceptions on the part of the group.
These later studies focus more on the communication taking place during writing, but still tend to
be preoccupied with measuring performance, rather than characterizing interaction.
There has been some work done which examines situated usage of collaborative writing
tools. An interesting but informal study that involved users working on a real task was the
27
evaluation of Interact, an early version of Aspects, by the Human Interface Group at Apple
Computer Inc. (Karimi and Palevich, 1990) . This study looked at two groups working to write
and edit a research paper and a meeting summary; the sessions were videotaped and participants
interviewed for their reactions to the use of the collaborative editor. All of the participants found
the experience of writing together engaging. Communication took several forms, including
external direct communication, messages typed in the body of the text, the use of the system’s
chat box, and indirect communication through the use of the telepointers and remote selections.
The system’s locking mechanism led to writing either using a scribe or by dividing the task up
into sections and later combining the text. The control mechanism became a problem during
close collaboration, when all writers wanted to access the text simultaneously, but some users
felt this helped maintain orderly collaboration. The lack of cues as to the location and actions of
others made it hard to tell what the rest of the group was doing, but direct communication helped
to overcome this problem. These results begin to suggest how a synchronous collaborative
editor could be redesigned to include enhanced awareness and communications support.
The impact of computer support for collaborative writing on social interaction has been
addressed in a number of studies of how computers can be used to help students write together.
For example, Davies (1989) described the experiences of a grade 4 teacher following the
introduction of computers and word processors into the group writing exercises in the classroom.
Compared to individual writing, group writing reduced writers block, increased motivation and
satisfaction with the final product, and improved the sense of audience. This suggests that the
social nature of group writing has a distinct impact on the character of the writing process.
Pontecorvo (1992) describes the impacts of collaborative computer use on writing in the
classroom. The researchers performed three studies of groups of students working together to
compose stories on shared computers. Results showed that the collaborative setting made the
ideation process an interactive process, with writers benefitting from the partner's competence.
Rather than being considered internally, ideas had to be expressed externally as the students
constructed solutions together. This expression of ideas can be divided along metalinguistic and
social dimensions. On the metalinguistic dimension, the ways in which the students proposed,
evaluated, defended and justified decisions could be divided into three levels. On the first level,
a student would propose an idea with no explanatory reflection, evaluation or justification. On
the second level, a proposal would be made with reference to existing framework constraints,
such as by evaluating previous propositions negatively. On the third level a student would give
overt semantic or syntactical reasons for a proposition. On the social dimension there is a
continuum that stretches between agreement, or co-construction, and opposition. In a co-
28
constructing situation proposals are juxtaposed, whereas in an opposing situation, ideas are
overly justified and argued strongly. Pontecorvo asserts that opposing interactions are more
productive for reasoning and learning processes. These interactions require reflection and the
production of sound arguments to support proposals, making the development of metalinguistic
competence a social, overt and explicit activity. The use of the computer facilitates interaction
and discussion, in terms of the shared screen, easy editing of text, and autonomous learning in a
socially supported situation.
These observations are supported by the findings of Daiute (1988) . Daiute describes a
study of 2 groups of 43 4th and 5th grade students who wrote 6 stories either alone or in pairs at
computers. The study investigated the conversations that take place during collaborative writing.
Data was obtained by audio taping all composing sessions, saving documents, and capturing a
computer keystroke record. The session transcripts were coded by 3 researchers, who developed
the categories for analysis as they went, then combined the codings where there was an 85%
agreement. The analysis also used content and subjective document quality analysis, participant
interviews, and keystroke logs. The findings showed that the students’ interactions were very
playful, and tended to continuously move forward rather than involving careful planning and
revising. They focused on negotiating the content of the stories, rather than on spelling and
punctuation. The researchers found that collaboration involves productive conflict, which forces
students to confront opposing views and resolve the conflict within the jointly written text.
McMahon and O’Neill (1992) studied collaborative creative writing using a system
called “bubble dialogues.” The goal of this system is the resolution of conflict by the negotiated
co-creation of “composite” characters; the students need to cooperate, debate, and hypothesize
the effects of their characters’ speech-acts, and then resolve any ensuing conflicts. Meaning is
created between people as they take up different positions in the developing discourse. This
study demonstrates the need to create mediating tools which not only respect this form of human
interaction, but also facilitate reflection on the whole process of creating meaning.
As with the observational studies of collaborative writing, these studies emphasize the
importance of social interaction and communication of ideas in the process of writing together.
The introduction of a shared text workspace provides a different way of viewing the developing
document, giving all participants an equal view of the text. However, there is more to the
process of writing than the formation of the text of the document. Systems need to support the
ideation process, the construction of the shared social reality in which the text is formed.
29
Chapter 2
The Prejudice ProjectMost previous studies of the use of collaborative writing tools have focused on the
mechanics of writing in short-term laboratory studies. I seek to address how synchronous
collaborative writing tools would work in a real situation, and how to use this experience to
inform the design of new systems. Writing together is not just a matter of choosing tools and
roles, and then moving through different stages. Rather, the process of writing together involves
coming to a shared understanding of the content and meaning of a text. Thus, the design of tools
to support collaborative writing is not just an interface issue. The designer must be sensitive to
the fact that she is designing a social artifact which will alter the social interactions of the group.
To explore requirements for the design of a synchronous collaborative writing system, the
designer needs to look at the extended use of such systems in a situated context. The Prejudice
Project is an attempt to do just that.
In this chapter, I will give an outline of the way in which the study was set up and
conducted, including a breakdown of how the students were introduced to the Aspects shared
editor. I will also present a detailed description of the groups that took part in the Prejudice
Project, and the setting and technology involved in the study.
The StudyThe Prejudice Project took place at the Huron Street Public School in Toronto between
January and May 1994. The study was undertaken by researchers from the Collaborative
Writing project of the Collaborative Multimedia Research Group, part of the University of
Toronto’s Dynamic Graphics Project (DGP). The project was carried out by myself and Ilona
Posner with assistance from staff at DGP and the Ontario Institute for Studies in Education
(OISE). Ben Smith-Lea worked as a technical assistant, taking notes and helping to run the
computers and video recording equipment. I also worked on the technical side, baby-sitting the
collaborative software and video equipment, and occasionally assisting with the students’ writing
activities. Ilona, acting as the instructor, led the discussions with the students and helped them
with their writing tasks.
The goal of the project was for grade six students to learn about prejudice while
collaboratively writing and producing a magazine on that subject. The magazine was to be
entered in a contest run by the B’nai Brith League for Human Rights, and used as part of the
“Whose Science Is It Anyway?” exhibit at the Ontario Science Centre. The magazine provided
an engaging task for the students to focus on, something which they could feel motivated to
30
complete, and about which they could develop a sense of ownership and involvement. It also
gave the students an extended task which would allow them to work together over a long period
of time, during which they could learn to work as a group, and become comfortable with the task
and technology.
We conducted an ethnographic study of two groups of students preparing the written
material for their magazine. Eight students were selected, with the assistance of their teachers,
from 14 volunteers out of two grade 5/6 classes. The students were experienced with the
Computer Supported Intentional Learning Environment (CSILE) shared knowledge building
system (Scardamalia and Bereiter, 1992) , but were not familiar with synchronous collaborative
work on a computer.
Each group met once a week for one hour after school over a period of twelve weeks, as
part of the school’s After-Four program. The students worked in one classroom, sitting at four
adjacent Macintosh computers (see Figure 2.1). The computers were connected by a local area
network, and all ran Aspects, allowing the sharing of documents among different machines.
Seating arrangements were rotated weekly to reduce the likelihood of subgroup formation and to
assign the preferred seats equally among the students. During the study they were given training
in the use of Aspects, introduced to collaborative writing techniques, and given the freedom to
use their new skills.
Figure 2.1: Physical setup of the Prejudice Project.
The first five weeks were highly structured in order to allow us to cover a variety of
topics related to prejudice and to introduce a number of collaborative writing techniques (see
Table 2.1). Students were shown the mechanisms provided by Aspects which were required to
31
perform the assigned writing tasks, such as working as a group to write a poem and a story. The
tasks were designed to expose the students to a variety of writing strategies (Posner, 1991;
Posner and Baecker, 1992) , and to show them the various ways of writing together.
32
Day Activity Date
1 Introduction to the project: goals, rules, responsibilities Feb. 1
Exercise: Differences: Interview a partner, getting to know each other better
Exercise: Similarities: As a group find what makes you all happy & sad
Exercise: Perception: Look at a picture write down 3 things you find most interesting
Goal of Exercises: Get students talking, expressing their opinions, more comfortable with each
other. See that they are all unique and have different opinions but that they agree about other
things.
Technical Goals: Start writing on-line.
Writing approaches: separate writers, scribe
2 Exercise: Perception: Advantages and disadvantages of differences in perception Feb. 8
Goal of Exercise: To show that people have different perceptions of the same things, people,
events.
Technical Goals: Introduction to groupware and group writing exercise, introduce concept of
revising their work.
Writing approaches: separate writers (in same document), joint.
3 Exercise: Minority Poem: When I'm in the Minority I feel ... Feb. 15
Goal of Exercise: To introduce concept of minority and how it might feel to be a member of a
minority.
Technical Goals: Introduction to Aspects.
Writing approach: separate writers (in same document), joint.
Exercise: Develop Orientation Plan: How to Welcome a New Student to Your Class.
Technical Goals: More practice with software and collaborative writing.
Writing approach: separate writers (in same document), joint.
4 Exercise: Generalizations: Need to qualify what we say, “Basketball players are tall” better
“most…”
Feb. 22
Exercise: Stereotypes: Discussion how stereotypes are created & group writing exercise.
Goal of Exercises: To introduce concept of generalization, stereotype, and prejudice. Make
them aware that all generalizations, particularly those about people, need to be qualified.
Technical Goals: More practice using Aspects and doing collaborative writing.
Writing approach: separate writers (in same document), joint.
5 Exercise: Prejudice: Define concepts, read a sample story, group writing exercise (story).
Technical Goals: More practice in collaborative writing.
Writing approaches: scribe, joint.
Mar 1
Table 2.1: Project plan for the first five weeks.
In the remaining weeks the students were free to choose what they wanted to work on and
how they would work together (see Table 2.2). The students' work during the second half of the
study included doing research, writing articles, creating artwork, and editing materials created in
the first five weeks which they had chosen to include in the magazine. After the initial training
33
in the uses of Aspects and the group writing approaches, we minimized our guidance so that we
could observe the choices the students would make. After the first three weeks, the students had
become comfortable with Aspects, and began to develop ways of working together in the shared
text document. These patterns of communication and social interaction form the basis for my
discussion and analysis.
34
Day Activity Date
6 Questionnaire #1 Mar 8
Magazine outline: Ideas about what should be done in the remaining weeks.
7 Magazine Work: Mar 22
Using printouts of all documents from first 6 weeks & Aspects
Group T: Survey questions, letter to Chretien
Group W: Glossary, editing existing documents
8 Magazine Work: Mar 29
Group T: Survey write up, use “Chat window” to pass the time, 5¢ for insults.
Group W: Survey questions, Word find, Current events, Collage, editing articles.
9 Magazine Work: Apr. 5
Group T: More survey write up, illustration for the Story, editing articles.
Group W: Survey questions, interview questions, editing articles.
10 Magazine Work: Apr. 12
Group T: Interview with League for Human Rights, editing articles.
Group 2: Interview with adviser, What happens around the world, editing articles.
11 Magazine Work: Apr. 20
Introduction, credits, deciding on the order of articles, Table of Contents.
Final push to finish everything up. Final editing.
12 Questionnaire #2 May 3
Preparing the Class Presentation
Individual Interviews
13 Class Presentation for Rooms 24 & 25 May 10
Table 2.2: Project plan for the remaining weeks.
The ParticipantsI will begin by presenting information about the composition of the two groups and the
relationships between the students in each group. I will also give some background on the
students, drawn from peer interviews conducted by the students in the first session, and from
casual observations conducted over the course of the study.
Composition of the Groups
The students who participated in the Prejudice Project were chosen from a group of 14
students suggested by their teachers. For convenience, I will refer to the group of students who
attended the Tuesday sessions as group T, and those who attended the Wednesday sessions will
be called group W. The students were drawn from two classes, rooms 24 and 25. Both of these
35
classes participate in the Ontario Institute for Studies in Education's CSILE project, which
involves extensive use of computers as part of the classroom curriculum.
Initially, a balance was maintained between the two groups with respect to the number of
students from each class and the number of boys and girls. However, because two of the
students decided not to participate after the first session, eventually both groups consisted of
three girls and one boy, and one group had three students from one class. This presented an
interesting imbalance in the group, leading to a certain amount of polarization in the interaction
between group members.
Group T
The members of group T were David, Elisa, Sam and Faith. David, a student from room
25, is an intelligent, self-confident boy, primarily interested in baseball and computers. David
often came to the sessions wearing a Blue Jays baseball cap, and talked at great length about the
World Series. David is very curious about computers, always eager to find out more about the
software being used and the efforts that went into developing the programs we had them use.
Faith, also from room 25, is a very bright girl, but is easily distracted from the task at
hand, and quick to criticize David. A not-too-friendly rivalry between David and Faith was
evident early in the study. This rivalry was based around their performance in class, and the fact
that David is eager to show his intelligence. As Faith said during one particularly
confrontational session, “David is so stubborn, he always decides everything for us. The teacher
asks what's one plus one, he says two without asking any of us, and even though its right it is still
annoying” [T5 54:39].2 David immediately followed this with “I'm the one who gets it right,
Faith” [T5 55:02], prompting an ugly look from Faith. David liked to control the group,
although he had difficulty maintaining that control. The rest of the group tended to criticize him.
He made much of this, claiming that “they all ganged up on me every time, cause I'm a boy.”
[T12 20:10]
The other two members of group T, Sam and Elisa, were both from room 24, and are
close friends. Initially, they both seemed to be very quiet, reserved children, especially Sam.
However, as they became more comfortable with the environment of the study, a mischievous
streak emerged. On their own they were both well behaved, but together they were very
rambunctious, giggling and goofing off at the slightest opportunity. A glance between them was
2References to the transcripts of the videotaped field study sessions will be given where appropriate; the notation
gives the group (T for Tuesday, W for Wednesday), the day number (i.e. 5), and the time at which the referencestarts (i.e. 45:27) based on the on-screen timecode.
36
often enough to send them into waves of laughter. Elisa is a very good writer, and expressed
interest in becoming an author; in one of the group activities, she stated that her one wish was to
write a good novel.
Group T, consisting of a pair of rivals and a pair of mischievous friends, was anything but
harmonious. Throughout the study, David battled with Faith for control of the group, a battle
which Sam and Elisa found hilarious, and occasionally very frustrating. The computers became
a focus of this conflict, enabling David to attempt to control the emerging dialogue within the
text being composed on the computer. Despite this conflict, the section of the magazine
produced by group T turned out well, and the students reported that they felt proud of their work.
They also claimed that they had managed to work together, and felt they had acted as a cohesive
group.
Group W
Group W, in contrast, appeared much more harmonious. The four students in this group
were Carla, Ryan, Sarah and Kate. Carla, Ryan and Sarah were all from room 25, whereas Kate
was from room 24. Carla, an awkward, often over-enthusiastic girl, has trouble with traditional
academic subjects, but is very artistic. Similarly, Ryan was rated as an underachiever by his
teacher. He is also extremely artistic, producing the cover for the magazine and offering to do
illustrations for both groups' stories. Ryan, like David, was the only boy in the group. However,
unlike David, he did not attempt to make an issue of this. He did occasionally attempt to control
the group, but rather than being treated with hostility he was usually ignored. Ryan's interest in
art was connected with an interest in computers; he told us in the first session that he wants to be
a computer graphics designer, and repeatedly asked us if we knew of any good art programs for
the Macintoshes. David and Ryan are friends; Ryan actually missed one session to attend
David's birthday party. This friendship formed a link between the groups, but also led to rivalry,
as David and Ryan compared each others’ groups progress.
Kate is a very quiet girl who preferred to work alone most of the time. In group
discussions she often remained silent, occasionally making unobtrusive comments that were
initially ignored but tended to get accepted later in the day. Sarah is also a reserved girl.
However, her shyness comes more from a sense of insecurity about her ideas than from a desire
to remain outside the group. She often presented her ideas as a question, and asked me and the
other researchers for reassurance. Interestingly, towards the end of the sessions she emerged as
the leader. She demonstrated an ability to organize and direct the group, a talent that helped to
lend her confidence when the rest of the group was faltering and lacking direction. Despite her
insecurity, she seemed to be the most mature student in the group, helping focus the group during
37
the final sessions and encouraging them to get the magazine written.
Group Organization
Both groups tended to avoid any explicit hierarchy or power structure. There were no
elected or appointed leaders, with decisions made for the most part by group consensus. Ilona
played the role of instructor and organizer. Early in the sessions, before the students became
familiar with the group and the tasks, they often appealed to Ilona for guidance and approval.
However, as time went on, they became more autonomous, only appealing to Ilona when they
were unable to resolve conflicts.
In group T, as has previously been mentioned, there were two dominant personalities,
David and Faith. In many of the writing activities, David attempted to gain control of the group,
an attempt which was assisted by the technology. Faith set herself up in opposition to David,
often opposing his ideas out of principle rather than because she necessarily disagreed with his
ideas. Sam and Elisa often backed up Faith in her criticism of David. The conflict was usually
resolved by David breaking off and working independently, Sam and Elisa forming a subgroup,
and Faith sulking and getting nothing done. The conflict between David and Faith came to
characterize the group's meetings. In the post-study interviews, David mentioned that one thing
he learned during the study was “how stressful it can be to work in a group” [T12 20:10], and
Sam said she enjoyed working on the magazine “except when David and Faith fought all the
time.” [T12 14:16]
Group W didn't initially develop any patterns of control, most of the students being fairly
moderate personalities. Ryan occasionally attempted to control the direction of writing
activities, but for the most part he was ignored by the girls, and would get frustrated. Kate
preferred to work alone; in group situations she tended to stay silent. Towards the end of the
sessions, Sarah emerged as a leader, possibly motivated by frustration with the lack of progress
demonstrated by the group. In the last few sessions, as the group worked to organize the
magazine, she took control, overseeing the rest of the group's activities and making sure that the
work got done.
The two groups worked independently on material for the magazine over the first half of
the study. After the midpoint, group W expressed an interest in working together. However,
group T, especially David and Faith, didn't want to merge, claiming that their group was better.
Despite the rivalry that existed between the two groups, there was also a common sense of
identity gradually developing, resulting from their privileged status as students who were able to
stay after school and work on the computers outside of the constraints of the CSILE project.
38
They also shared the common goal of the Prejudice Project, motivated by the League for Human
Rights contest and the Ontario Science Centre exhibit. This common identity was expressed
most clearly when they presented the results of their work to the combined rooms 24 and 25
classes after the end of the sessions.
Setting: The ClassroomsThe two classrooms are situated on the third floor of the old Huron schoolhouse, a small
building adjacent to the main school building. Due to the proximity of the classrooms, and the
fact that both classes are part of the CSILE project, the students tend to know each other.
The two classes used are both grade 5/6 splits, and are alpha-test classes for the CSILE
project. This involves the use of CSILE, a computer program designed to facilitate collaborative
knowledge building in the classroom as an integrated part of the curriculum. Both classes are
well equipped with computers, each containing approximately 10 Macintosh IIci computers
connected by a network to two file servers. The computers were installed as part of the Apple-
sponsored CSILE project approximately 3 years before the Prejudice Project, and at the time
were considered powerful. The teachers have both integrated the computers, to varying degrees,
into the daily curriculum. Despite this, however, there is a visible distinction between the
computers and the rest of the classroom, as can be seen from a detailed look at the layout of
room 25 (see figure 2.2).
39
Figure 2.2: Floor plan for Room 25, the site of the Prejudice Project.
The students' desks are arranged in pods around the class, with groups of 4-6 students
sitting at hexagonal clusters of desks. Each student has a yellow bucket to hold books, papers,
pens and other belongings. Students in each cluster are responsible for keeping the area clean,
and for stacking the chairs at the end of the day. Books from the bookshelves around the
classroom are scattered around the desks. The walls to the left and right of the doorway are hung
with projects and paintings done by the students, and art projects are hung from the ceiling.
In contrast to this, the computer stations are set up in two locations, along the far wall and
in a back room. The computers in the back room are arranged along two walls, facing away
from the door. The computers in the main class are in a row along one wall, facing the windows.
Both computer locations are completely devoid of decorations, showing no personalization or
sense of ownership on the part of the students. The students are able to use any of the networked
40
computers, and usually do so only during set periods dedicated to CSILE work. The computers
run a restricted version of the Macintosh operating system, called At Ease, which provides a
subset of functionality and prevents the users from tampering with the system. A password is
needed to get to the Finder and full system access. In addition to this, entry into CSILE is
controlled by a login and password given to each student. Each day, when we were setting up
the video equipment, I would notice the day's schedule written on the board, with the limited
times dedicated to computer use included in the daily activities. These restrictions on the use of
the computers, and the lack of personalization of the computer workstations, suggests that the
computers had not become an integral part of the classroom society. The teacher's desk is at the
head of the room, raised a few inches above the rest of the class. The teacher has a more
powerful computer, which does not use the At Ease system. The teacher also has a laser printer
which the students can use to print out projects.
For our study, we used the computers in the main classroom. For the majority of the
activities, the students sat at the computers to first discuss their work and then write on the
machines. This actually proved to be a problem, as it was very difficult to stop the students from
immediately jumping on to the computers without any discussion. Later in the study, as the
students worked to organize and edit the magazine, they tended to shift between working on the
computers and coming back to their desks and working on paper.
Our recording equipment added another dimension to the physical setting of the
classroom. Since we were interested in the students' use of the computers, we set up a bank of
cameras facing the machines. Two cameras were focuses on the screens, and two were used to
capture the students' interactions. Behind the cameras, we set up a rack of VCRs and an audio
mixer, with a video monitor and headphones. We also had a laptop computer set up to take field
notes. Ben and I sat behind this equipment, and Ilona moved about in front, coordinating the
students' activities.
This created a divided space, with the front stage centred on the students, and the back
stage behind the cameras. Although we were supposed to be invisible behind the cameras, Ben
and I often became involved with the activities. The students drew us in, either asking for advice
on their work or as part of their informal interactions. Thus, the divisions between “subjects”
and researchers was not a firm line, and often shifted back and forth.
The TechnologyThe students in rooms 24 and 25 are part of the CSILE project run by OISE. CSILE
(Scardamalia and Bereiter, 1992) is a shared database that the class uses to build up a communal
41
knowledge base to be used as part of ongoing research and assignments. Students contribute to
the database by entering text or graphical “notes” which can be read by other students. They can
also add comments or material to other students' notes, and make links between notes. CSILE is
intended to support intentional learning, knowledge building, and collaboration. The use of
CSILE in everyday classroom activities makes the students familiar with using computers. It
also adds a dimension of computer-mediated communication to the social fabric of the
classroom. This familiarity with the use of the computer as a medium for communication and
social discourse had an impact on the development of the group in the Prejudice Project.
The computers played an important role in the motivation behind at least the boys'
involvement in the Prejudice Project. Ryan told us on the first day of the study that he hopes to
be a computer graphics designer. He was especially intrigued when I told him I work in the
Graphics Lab at the University. David was similarly fascinated by the computers. For David,
the computer was an object to be controlled, and to use as a means of control. One of the first
things he did when the students started working on the computers was discover that Ben was
listed as one of the contributors in the “About...” menu for CSILE. When he discovered this, he
confronted Ben, repeatedly calling him “Benjamin Smith-Lee” without having previously been
told Ben's full name. This surprised Ben at first. He looked puzzled that David knew his name,
and asked him how he knew it. David revelled in the power this information gave him, at first
refusing to tell Ben how he had discovered his name. David was also extremely curious about
the development of the software we were using, asking me how long it took to write the
programs and how hard they had been to develop. David's proficiency and understanding of the
programs enabled him to use the computers to attempt to gain control over the group. The
technology became a focus of much of the group's work and interactions.
By the end of the sessions, the two groups had come to identify themselves as a distinct
group, the “After-Four Group”, using this as a status symbol over other classmates. This status
was centred around the technology involved in the study. Each Tuesday and Wednesday, the
other students would see us come bustling in just after the 3:30 bell, and watch in awe as we set
up all the video and audio equipment. They would often want to play with the cameras, and ask
to see the images on the video monitors. The fact that they were involved in this unusual
technological endeavour gave the group a special status.
The Aspects Collaborative Editor
The computer program used by the students during the Prejudice Project, Aspects, is a
42
collaborative editor, a word processor that allows students to work on a single text shared
amongst the entire group on separate computers (see Figure 2.3). This common text creates a
shared task space through which the group can communicate and interact. The text that appears
on each person's screen is the same. As each user makes a change on her computer, the changes
show up on all the other users' screens. Each person can be looking at a different part of the text,
but the overall document will be the same. Remote users’ selections show up as hollow
rectangles, whereas the local user’s selection appears in the usual manner, as a reversed
rectangle.
Figure 2.3: The Aspects shared editor.
Aspects also provides a variety of communications tools, including a chat box which lets
students communicate informally, and telepointers, cursors which appear on other people's
screens. Although the tasks given to the students in our study involved using Aspects to work on
writing activities, the computers were used for much more than these task-focused activities.
The fact that the shared document is often the focus of group interaction makes it an
important means of negotiating for control and shared understanding within the group. To
understand this, it is helpful to explain some of the features of the Aspects system which play a
role in these processes. Because of the nature of the technical problem of distributing a text
document across a computer network between several users, and maintaining consistency
43
between the different copies of the document, a number of control mechanisms have to be built
into the system. There are several ways to approach this, each of which has technical and, as I
will describe, social consequences.
In Aspects, all users are able to enter text at the same time; however, only one person
may be editing a given paragraph at a given time. The bar down the left of the text window
indicates whether someone has the paragraph locked. The black bar is the local user, whereas
the grey bar is a remote user. This form of locking mechanism tends to encourage certain types
of access and control within the writing task. For example, if a document consists of a number
of paragraphs, and people are working on separate paragraphs, then everyone can work at the
same time. However, if the discussion is tightly focused on one paragraph, or the document is
made up of only one paragraph, then one person is able to control the content of the document.
If the group interaction is centered around the document, this gives the person with control over
the document effective control of the entire discourse. Others are only able to directly contribute
if the controlling person releases the paragraph.
This shifting of control to the document moves the focus of the discourse from people
and ideas to the concrete representations of the words on the computer screen. No longer are the
ideas just abstract notions; instead they are almost physical objects which can be manipulated.
This changes the types of skills needed to successfully get ideas accepted. Rather than just social
skills, a person also needs to be comfortable with the technology and to have spatial and
mechanical skills. However, in our groups the shift is not entirely from social to technological,
since the group members are all working face to face. Thus, to succeed in a negotiation for
meaning or control, a group member needs to use both social and mechanical skills. This is a
more complex task than traditional face-to-face social interaction.
45
Chapter 3
Data Collection and AnalysisIn this chapter, I will discuss my choice of methodology and approach to data analysis,
and how this choice relates to the approaches taken in other studies.
Data CollectionTo study the social interactions taking place during collaborative writing, we needed to
capture as much information as possible about the activities of the groups. The various types of
information collected during the study had different uses, many of which were not immediately
apparent at the beginning of the study. As with any complicated project, the actual setup altered
considerably during the course of the project. I will now discuss each of the forms of data
collection, the purpose behind it, and how the collection method altered during the course of the
study.
Audio and Video Recordings
During each weekly session all of the group’s interactions were videotaped. The
recording setup included two cameras covering the subjects working on the computers, and two
cameras capturing the screen images (see Figure 3.1). All subjects wore lapel microphones
which were mixed onto the left and right channels of the videotapes. Audio was also recorded
on a four-track audio tape.
Cameras Cameras
Workstations
1 2 3 4
Figure 3.1: Data collection setup.The students sat at positions 1 through 4.
Initially we planned to mix the audio with a four way split screen of the people and
screen shots to give us a single tape for each session. However, after the first few weeks the
usefulness of this approach was no longer apparent. For simplicity, we abandoned the four-track
46
audio, and put the sound directly onto all video tapes, providing consistent audio across all the
images. This was very useful for locating interesting incidents on the various tapes, and reduced
the complexity of both the recording process and the setup needed for analysis.
Transcriptions
Exploratory analysis was conducted on the video records of the sessions by Ilona and me
using the Timelines (Harrison, Owen and Baecker, 1994) sequential data analysis and video
annotation software. In addition, transcriptions were made using Timelines while the sessions
were in progress. These three transcriptions were subsequently used in the analysis of the
sessions. This analysis focused on identifying common problems, incidents and trends in the
data. Although Ilona and I had access to all the tapes from each session, I found the people tapes
more useful, and tended to focus on the interaction between subjects, whereas Ilona focused
more on the events on the computer screens. This difference in approach led to a broader range
of information being derived from the tapes. Once the data had been transcribed and coded for
interesting events and trends, the transcripts were merged, and I re-analyzed the tapes to add
detail and accuracy. More in-depth discourse analysis and coding was then carried out on
selected portions of the data to explore in detail the notions of control, attention and grounding of
communication in the shared workspace. The initial analysis was used to identify the areas of
interest and provide an indexing for the video data.
Teacher Evaluations
The teachers of the two classrooms being used in the study were asked to evaluate the
students' writing skills and ability to work together. Each individual's writing skills were rated
using a seven point scale according to vocabulary, spelling, grammar, ideas and creativity, and
writing style. The teachers also provided blind evaluations of a selection of work, both
individual and group, using these categories. These scores were then averaged for each
individual, and across each group, to give an indication of general writing level. Each student's
ability to work in a group was also rated according to interpersonal skills, communication skills,
leadership skills, maturity, and task focus. This material will not be used in this thesis, but wasincluded in (Posner, Mitchell and Baecker, in press) .
Physical Artifacts
Printouts of the documents in progress were obtained at various points throughout the
sessions. This provided valuable insight into the content of the group work being performed by
the subjects. A copy of three of the documents produced by each group, the introduction to the
magazine, a poem about how it feels being a minority, and a story about prejudice, are included
in Appendix B. These documents are refered to throughout the analysis.
47
Journal Entries
At the end of each session, we asked the students to write a short journal entry, in the
form of a “What I Learned Today” note in the CSILE database. These entries were very useful
in the analysis of the data, often providing us with much different insights into the group
interactions than were available by simply observing the videotaped sessions. This suggests the
importance of subject participation in the analysis of complex social situations, participation
which we were able to approximate through the journals, questionnaires and interviews.
Questionnaires
Questionnaires were given to the students at the midpoint and the end of the study. In
hindsight a pre-study questionnaire might have been useful. The questionnaires gave us insight
into the attitudes of the students, which were often at odds with the behaviour we observed
during the sessions. For example, after writing a story on day 5, the Tuesday group had an
argument about who wrote the story, with each group member claiming ownership. There was
also a fight over the placement of a comma in the last sentence. However, despite this conflict,
Faith wrote in her questionnaire that she felt they had learned to work together and had written a
story as a group.
Final Interview and Group Discussion
Along with the post-study questionnaire, a final interview was conducted with each
student to ask more open-ended questions about the study. Finally, a few weeks after the end of
the project we came back and conducted a group discussion about the project, which gave
additional insight into the students' experiences. Both the final interviews and the group
discussion were videotaped and transcribed, and formed part of the data used for later analysis.
Approach to the Analysis of the DataWe analyzed the collected data from several perspectives, looking at the technological
and interface requirements, and the impact of the technology on both the students' process of
learning to write together and their social environment. The issue which I faced was what
approach to use to evaluate and analyze the data. The domain of research could be viewed as a
complex problem-solving task, which suggests a quantitative approach, but, as our survey of the
literature has suggested, the social interaction and communications aspects of collaborative
writing are also important, and suggest a more qualitative approach. I chose to use an
ethnographic approach, viewing the video data as rich descriptions of the group processes and
interactions. The tools I used to analyze this data were drawn from ethnography,
ethnomethodology, and interaction analysis.
48
Ethnography is defined by Suchman and Trigg (1991) as the “careful study of activities
and relation between humans in a complex social setting.” (p. 210) The discipline is based on
extended participant observation of the internal life a of setting, in order to understand what the
participants take to be the relevant aspects of their activities. Ethnomethodology is the attempt to
“locate meaningful realities of everyday life in action, in the linkage between structures, things
and behaviours.” (Gubrium 1988, p. 16) Interaction analysis is the detailed investigation of the
interaction of people with each other, and with their material environment (Suchman and Trigg,
1991) .
These disciplines have much to offer the study of groupware systems. My goal is to
incorporate the approaches taken by these disciplines in an attempt to focus on the group
interaction and the use made by the group of supporting technology and artifacts. I will now
discuss some of the issues involved in this choice of approach to the data, drawing on the
approaches taken by other researchers to provide background and justification for my choices.
Approaches to Studying Writing and Collaboration
To give the problem some context, it is useful to look at approaches taken by other
researchers. Similar methodological issues are involved in the study of individual writing.
Fitzgerald (1992) examines the different methods of research used in the study of writing, as
they are related to the different views of writing and knowing: the exogenic approach, which
searches for an observable, objective truth and tends towards logical positivistism, vs. the
endogenic approach, which has a more internal view of knowledge, and takes a social-
interactionist attitude to research. She observes that the researcher’s “problem definition, values,
and beliefs about ways of knowing interact forcefully with the choice of methodology” (p. 43).
Problems tend to be defined to fit methods, and similarly methods tend to shape problems.
Writing researchers who are concerned with content and product have traditionally tried
to classify revisions using content coding. Those concerned with process, such as Hayes and
Flower (1989) , use tracing and think-aloud protocols and simulation by intervention in an
attempt to access the writer’s internal cognitions. Researchers more concerned with the social
aspects of writing tend to use participant observation, collecting anecdotes, samples of work, and
rich descriptions of the experience of writing in a community of readers and writers. There is no
one correct methodology; in fact, she argues that a multiplicity of methods is beneficial. There is
a need to temper different points of view, to encourage methods that will complement each other,
rather than cause conflict. In my study, the focus of interest is the social interactions that
occurred within the groups during writing, suggesting an endogenic, ethnographic approach.
49
The current dominant paradigms in the study of computer-mediated communication for
collaboration, according to Mason (1992) , are the postpositivist, interpretivist, and critical
theory approaches. Postpositivism is mainly quantitative; interpretivism focuses on the
qualitative, and is based on time and place; and critical theory tends to be a historical, political,
and structural based approach. Various evaluation techniques have been advocated by different
researchers. Questionnaires involve the use of pre- and post-study questionnaires, investigating
the users’ expectations, the characteristics of users, and their perceptions of a system. The
disadvantages include the fact that the wording of questions can be critical, questions can be
biased, and it isn’t clear whether open or closed ended questions are preferable. Interviews take
a grass-roots, close-contact approach, but are difficult to do well, and the researcher needs
experience. Empirical evaluation, the approach often favoured in the CSCW community, is
more controlled, but ignores the impacts of long-term use, context, and the feeling of ownership
of the system on the part of the users. The use of statistics, involving extensive coding of
communications, can mistake activity for learning, taking no consideration of message content.
Participant journals are an option, making use of self-reporting information, with the attendant
biases. Content analysis can be useful, but is time-consuming. The advantage is the ability to
use both quantitative and qualitative analysis, leading to results that are both detailed and
generalizable. The case-study approach is inductive, making use of participant observation
which leads to detailed, situated observations.
An Ethnographic, Case-Study Approach to Studying Technology
Our aim is to look for the human interactions involved in the use of collaborative
software, and analyze how the artifacts fit into this interaction. From this, it is possible to
develop theories about the politics surrounding the artifacts, and the role of the groupware
system in the human interaction. Due to the situated, social nature of the task of collaborative
writing, and the extent to which it is inextricably interconnected with the context and
environment in which the activity is taking place, I have adopted a case study approach to the
study of the students working in the Prejudice Project.
According to Wagonner (1992) , the use of a case study approach allows for both
quantitative and qualitative analysis. It makes use of a real life context, examining individual
cases in a detailed fashion. The “key to the explanatory power of the case study is the systematic
presentation of a cogent and compelling argument, supported by credible evidence, that
demonstrates a consideration of the alternative points of view and explanations of facts and
inconsistencies.” The case study offers a comprehensive approach to understanding the complex
process of collaborative work using computer systems.
50
Our study is what McGrath (1993) would call a field experiment, similar to a field study
in that there is an effort to make direct observations of groups in an existing context, but with the
difference that there is one major intervention, in this case the introduction of Aspects, the effect
of which we intended to study. This attempts to maximize realism of context, while at the same
time allowing us a certain amount of precision in control and measurement of behaviour. Our
study can complement other approaches, such as the ShrEdit studies (McLaughlin Hymes and
Olson, 1992; Olson, Olson and Meader, 1995) , which are more concerned with generalizability
over a population and control of variables at the expense of lack of realism. Realistic
observation of interaction between computers and users is a challenging task. This challenge is
increased when the software is extended to support group work. As discussed by Grudin
[](1989, p. 176-7),
it is difficult to create a lab situation that accurately reflects the social,
motivational, economic, and political dynamics that are quite likely to affect theuse of groupware. . . . Evaluating groupware “in the field” is remarkably complexbecause of the number of people to observe at each site, the wide variability of
group composition, and the range of environmental factors that play roles indetermining acceptance [of the system] . . . [determining] success or failure willbe easier than establishing which underlying factors brought it about.
It is important to realize that the context in which the group is working directly affects the
conclusions drawn from the study. In fact, Fulk, Schmitz and Schwarz (1992) suggests that one
should focus on the specific context of interaction when developing a theory about groupware
usage, rather than attempting to assert generalizable hypotheses. The constantly shifting social
context forms an integral part of the use of the computer tools, a context which must be reflected
in the system design.
Using Ethnography to Inform Design
A major difficulty in the use of ethnographic data is the translation of ethnographic
descriptions into design requirements (Fafchamps, 1991) . Shapiro (1994) discusses the limits
and challenges of using an ethnographic approach to inform system design. Ethnography, as
used in CSCW, is more accurately described as ethnomethodological ethnography;
ethnomethodology sets for itself a strict agenda which insists on a rigorously descriptive rather
than an theoretical or explanatory program. This lends it strength in producing rich descriptions
of work in context. It is the strictly documentary method which is the basis for the claim that
these rich descriptions can be seen as systematic, not arbitrary. However, it is rare that
ethnomethodological studies adhere strictly to descriptive material, with supposedly faithful
reporting easily slipping into the explicative and the interpretative. In fact, it is in this move
51
from strict adherence to the material into the development of theoretical structures to describe the
material that sociology becomes most interesting. The choice of whether or not to maintain a
descriptive approach is most important when attempting to use ethnography to inform design.
System design is by definition concerned with categorization and an attempt to theorize and
predict behaviour, so that a system can be designed such that it matches the expected behaviour.
The descriptive material provided by ethnography can be used to inform the design of systems.
The problem with using descriptive ethnographic material to inform design is that
designers tend to look for broad, generalizable characterizations of a user population, the sort of
results usually obtained through controlled experimentation. Ethnographic investigation
produces “a contextualized reproduction and interpretation of the stories told by the subjects.”
(Vidich and Lyman, 1994, p. 43) The underlying assumption is that there is no one truth or
reality, and that all understanding must take place within a given context. There have been many
attempts to overcome the seeming contradictions between this basis and the desire to generalize
results. Glaser and Strauss (1974) claim that theory should emerge from the field, with the
gradual development of a corpus of data leading to the emergence of a grounded theory of the
area of interest. This can be contrasted with the extended case method, which takes the approach
that, “rather than theory emerging from the field, what is interesting in the field emerges from
our theory.” (Buroway, Burton, Ferguson, Fox, Gamson, Gartrell, Hurst, Kurzman, Salzinger,
Schiffman and Ui, 1991, p. 9) I have tried to strike a balance between these two approaches.
Beginning with a grounding in the theory of collaboration and writing, I tried to leave this theory
behind and let the data stand on its own, but then pulled the observations derived from the data
back into the theory to draw out the design requirements. This recognizes that design takes place
in a specific context, and builds on existing understanding of the domain. Rather than offering
generalized conclusions, the design requirements are descriptive, adding to the body of material
that characterizes the ways in which people write together.
For my analysis I am attempting to incorporate all of the information at my disposal,
including my field observations of the students, drawing on the impressions I developed having
got to know them over the 12 weeks of the study. I also made extensive use of the video tapes,
using them to gain insight into the interpersonal interactions and the activities performed by the
students. They were also used for more detailed analysis, using approaches such as conversation
analysis and a study of the students' body language, specifically gaze direction. I had access to
the documents created by the students, having saved these at various stages of their composition,
including the final magazine submitted to the League for Human Rights. I was also able to make
52
use of the interviews, questionnaires, and final group discussion, which focused fairly
specifically on the task of collaborative writing and the way the technology was used in this
process. The type of material available to me for analysis included a rich set of different views
of the groups’ work. However, the most useful data, which I used as the focus of much of my
analysis, was the videotapes of the students’ interactions. The videotapes were used as a
documentary recording of the events to be studied; as the transcripts were developed, a number
of concepts began to emerge from the data.
The initial analysis consisted of transcribing the video tapes and attempting to code them
for interesting events, where interesting is defined very loosely to include anything relevant to
the use of the technology, collaborative writing, and conflict resolution. In the initial pass
through the data, Ilona and I did not make any attempt to take into consideration the group
dynamics, the social setting, or the interpersonal relationships present in the groups. Then, in an
iterative process, I went back and tried to build up a set of incidents which formed a
characterization of the students’ use of Aspects to write together.
In combination with the interpretations and textual descriptions of the interactions which
I built up during my analysis, the video can be considered to be a form of thick descriptions of
the groups’ activities during their participation in the study (Goldman-Segall, 1991) . These
descriptions were gradually grounded in the theoretical background. The difficulty with using
this form of data as the central component of my analysis was the question of how to convey the
observations rooted in the video data to the reader of a textual thesis in a rich meaningful way. I
have attempted to approach this problem by including as much rich description and direct
examples from the transcripts as possible. It would be interesting to explore how to take more
direct advantage of the richness of the video data. The video could be used to add detail to the
transcriptions, allowing the reader to make her own interpretations of the interactions. The
challenge is to present the video data in context, with enough detail to make the video segments
meaningful to the reader.
53
Part II: Analysis of Writing in a Shared Workspace
The Prejudice Project was conducted to investigate how the use of a synchronous
collaborative editor affected the social interactions which take place between group members
while writing together. In this section, I discuss the results of my observation and analysis of the
students’ interactions during their work on the magazine “Prejudice in the 90’s” (see Appendix
B).
The introduction of the shared workspace provided by Aspects altered the ways in which
group members were able to negotiate for the control of the discourse during writing. The access
control mechanisms embedded in the shared workspace were used as additional means for
controlling the discussion, leading to new patterns of control and negotiation.
The shared workspace also altered the ways in which the group members were able to
compete for and allocate attention. For the Wednesday group, the text provided an alternative
focus of attention, allowing group members to avoid giving attention to others; for the Tuesday
group, the shared workspace came to represent a primary focus of attention, often used as a
means of gaining attention and control.
The shared workspace therefore came to play an important role in the grounding of
shared understanding between the group members. I discuss the costs and affordances of this
new communication medium, and how these costs and affordances change the nature of
communication and collaboration.
Finally, I look at the approaches used by the students to synchronous collaborative
writing, examining how these differ from other forms of collaborative writing, and how they
relate to the models of individual writing discussed in the previous section.
55
Chapter 4
Patterns of Control Using a Shared EditorI will begin my analysis by looking at the nature of communication in the two groups
which took part in the Prejudice Project, focusing on the members’ negotiation for control and
meaning within their discourse. I will examine how the use of computers, and specifically the
Aspects networked editor, influenced the flow of control and negotiation of ideas within the
group. The way this happens depends very much on the individuals in the group, and the
formation of various subgroups. The technology changes the way ideas are communicated,
requiring different skills for success.
Aspects provides a shared workspace in which several writers can work at the same time.
The most natural way for new users to make use of this shared workspace is to have one group
member act as scribe, while the rest of the group follows along and makes contributions. This is
similar to how a group would work with a single computer and a single-user word processor.
This is how the groups in the Prejudice Project tended to work initially. Aspects’ use of a
paragraph-level locking mechanism affords this type of writing, since only one person can
actively make changes in a given paragraph at one time.
The scribe has significant control over the group. Group members are able to see the
text, but ideas have to be approved by the scribe before she enters them. The scribe has control
of the focus of the discourse, and thus has the responsibility for negotiating resolution of conflict.
There can be a number of different approaches to control on the part of the scribe. How
effectively the scribe is able to deal with conflict between the ideas being entered and the various
ideas of the group members influences the success of the writing task. The scribe can silence the
group members, using her position of control as a way to suppress conflict, or can direct the
group members, acting as a moderator and attempting to decide which ideas get accepted.
Silencing is effective to a point, as long as the scribe has the ability to maintain social control of
the group as well as control within the text. However, denying group members expression of
their ideas is bound to lead to tension and conflict. Directing the group and maintaining the flow
of ideas is more conducive to resolution of conflict, but requires social skills beyond the simple
retention of control through the use of the paragraph locking mechanism.
Aspects does allow several people to enter text simultaneously; however, given that the
scribe has control of the focus of discourse, it is difficult for group members to shift control of
the floor or to start a parallel discourse. It is possible, though, which leads to the potential for
parallel writing (see chapter 7), where several people are entering ideas with a common
56
discourse focus. If the scribe works closely with the group members, or frequently allows
control to be shifted, it is also possible to approximate joint writing.
I will now look at how the various group members used different techniques to attempt to
gain control of the discourse, and how well these techniques worked depending upon whether the
discourse was located on or off the computers.
Control Away From the ComputerIn the Wednesday group, there were no particularly dominant personalities; this led to
frequent disputes for control between the group members, with control shifting back and forth
depending on the task and nature of the discourse. The most interesting instances of control and
the role of the computers in the negotiation of control occurred between Ryan and Sarah. In
social situations, Ryan often had trouble getting his ideas across to the group. For example,
when the group was working around one of the desks in the class editing a group-written poem
(see Appendix B), Ryan had difficulty getting his suggestions acknowledged:3
Ryan when I'm in the minority I feel, first we should havewhat everyone wrote then at the end everyone shouldcontribute to a helpful thing
Kate make sure that they are not excluded [to Carla]Carla I said included [to Kate]Ryan its in poetic form. everyone is kind of just doing
things. so you have all your entries down and thereader will think well how can that be helpful?
Ilona do you want to actually do what you are trying tosay, actually write, maybe the writing would beeasier than the explanation
Ryan everyone would add in what they think would help theproblem. Let me explain it one more time. At the endof each
Carla you wrote this one [to Kate][W8 23:06]
Ryan made repeated attempts, starting and restarting his utterances, then stopping in frustration.
The others continued to discuss a different issue, debating whether a certain word should read
"excluded" or "included", ignoring Ryan's contributions. When he doesn't get the floor
immediately, Ryan gets annoyed and tends to give up, assuming that no-one is listening to him.
His style of gaining the floor is not forceful enough to give him the immediate results he expects,
nor is he persistent enough to eventually gain control of the interaction. In fact, the rest of the
group also tends to get frustrated, and occasionally they will make fun of Ryan for his failure to
3In this chapter I am using a simplified transcript notation. Comments are in square brackets, and all other text is
spoken. Punctuation is used to denote intonation, and italics are used to denote emphasis.
57
gain control, further undermining his attempts. Ryan tends to try to use a silencing approach to
control, but does not have the confidence to carry this out away from the computer.
Ryan’s difficulty gaining control of the direction of the group discourse is further
frustrated by Sarah. She is usually fairly silent, and when she does speak she often frames her
ideas as a question, revealing her insecurity about her statements. However, she shows a definite
ability to organize and direct the group, especially when this means she does not have to
contribute her own ideas to the interaction. For example, during the editing of the magazine, she
made a lot of the executive decisions with respect to corrections and assignment of tasks [W10
14:28], including checking up on other group members' progress [W10 39:44]. On the last day,
she took charge of organizing the order of the magazine:
Sarah perception goes third, when I'm in the minority, Ifeel, I feel what? OK lets work with me here, OKwe've organized it so that it goes intro, perception,minority I feel
[W11 47:15]
She feels more comfortable when she does not need to rely on the other group members for
direction, preferring to order and control the others' ideas. This preference for the directing style
of control turns out to be well suited to the use of Aspects.
Control in the Shared Text WorkspaceThe introduction of technology into the group subtly alters the dynamics of the group.
Using Aspects, Ryan is able to gain initial control, since he is able to use the paragraph level
locking mechanism to grab control of the text, and thus the focus of the discussion. He is able to
get the immediate control which he is denied in a social setting. A good example of this
occurred while the group was composing a story during the fifth day (see Appendix B). The
group is discussing the next paragraph in the story. Ryan is trying to get his idea accepted, but is
having trouble getting the two girls to listen to him. They try to enter their ideas, but Ryan has
control over the end of the document, the current focus of the group’s discussion, and is able to
retain control through the computer program.
Ryan well, the, this girl stands up and says all homelesspeople aren't bad
Kate stands up andRyan and the kid says well a homeless person is just
someone who doesn't have money or just enough forfood and doesn't
Kate or doesn't have enough money to buy a houseRyan and they still have a personality and stuffKate [starts typing]Carla [starts typing]
58
Ryan hello what are you doing? Carla I'm trying to typehere
Ryan okay, what's his name [starts typing]Kate I think that's a little bit far downRyan okay, soCarla one day, whenRyan [looks over at Carla] Carla, what? I'm typing itgroup [silence, Carla grins]Kate [leans over] unless you areRyan okay, so he sees like a homelessKate [adds something]Ryan Carla, stop! tell her to stop, tell her not to spoil,
so when he saw [types] the scruffy looking old manCarla scruffy, what's that meanRyan so he sees him talking to a wallKate scruffy, that'sCarla scruffyKate maybe drunk?Ryan so tell me, and you can go on, we can fill in details
laterCarla what's scruffy mean?Ryan it means like dirty - okay tell me what to say, like,
next, okay [Ryan typing, no longer provides ideas][W5 18:15]
When Sarah is not there, this strategy works successfully. Having gained control, Ryan takes on
the role of the scribe, a purely mechanical task which does not involve social skills. From this
point on, he stops contributing ideas. Carla, who is very good at composing and contributing
ideas but lacks the mechanical skills required to enter them, provides the content and Ryan
transfers them to the computer.
Despite this smooth collaboration, there are moments of frustration on Carla's part, as can
be seen later in the same session. She keeps offering ideas, but Ryan replaces them with his
own:
Carla the next day theyRyan no, yeah they're gonna talk to himCarla or buy him even something to eatRyan yeah, the next day [typing] maybe it should be like
Christmas or something and they give him somethingCarla yeahRyan the next day it was no, okay one sec, okay the next
day [typing] come to see oldRyan the next day they are going to talk to himCarla they are going to talk to him and maybe buy him
something to eatRyan maybe it was Xmas and they are going to give him
something No it was Valentines day
59
Kate it was Valentines dayCarla that's betterKate the next day they came back to seeRyan the old man and [typing]Carla and before they knew itRyan they were having a good timeCarla talking, eatingRyan talkingCarla and laughingKate talking and laughing with the old manRyan with [typing] okay
[W5 35:00]
Ryan, as the scribe, is maintaining control of the focus of the group's discussion. Although ideas
can be negotiated between the group members before being entered, Ryan acts as a filter, having
the final say as to what gets on the screen. Once text is visible on the screen, Ryan also has
control over what is altered. Any requests for changes must go through him. Carla must keep
repeating her ideas until Ryan acknowledges them and adds them to the developing text,
otherwise they will be lost.
This works fine as long as there is no cognitive conflict. However, if the ideas on the
screen do not match with the ideas being discussed amongst the group members, the scribe, Ryan
in this case, must make use of social skills to maintain control of the discourse. This is
something Ryan is not able to do. When there is a need for negotiation, Ryan tends to get
frustrated, and occasionally loses control of the discourse. This happened during the final
session, while the group was working to write an introduction to the magazine (see Appendix B):
Carla we just also should know that its saying minoritiesand
Ryan no its an introductionCarla its an introduction [getting annoyed at Ryan] how
about a table of contents then?Ryan at the back you see an about the author that's where
you see well and Ryan is 12 years old and theintroduction is what's happening in the magazineCarla get that hand [telepointer] out of there please
Sarah what happened in the magazine we're explainingRyan we interviewed different people of different ages and
recorded their thoughts about prejudice acts. thisall links up to our
Sarah all the topics we'll be covering link up to our topicprejudice in the 90s I dunno what do you think
Ryan all the topics, I dunno, migraines, migraines [holdshis head]
[W11 25:20]
Ryan is ignoring the contributions of the other group members, trying to silence them by
60
repeating his ideas. They do not respond, having conflicting ideas but no means to add them to
the text. At this point Ryan must relinquish control of the paragraph, otherwise the discussion
will come to a standstill. What is needed is direction, organization and negotiation, skills Ryan
lacks.
Sarah, on the other hand, is able to organize the ideas being thrown out by the group
members, manipulating the verbal discourse in a way not available to Ryan. However, she does
not initially have the confidence with the technology necessary to carry this control across to the
computer. In a situation where the focus is on the on-line text, control tends to move back and
forth between Ryan and Sarah. Especially early in the study, neither Ryan nor Sarah were
confident enough to take complete control, as can be seen in the following excerpt. The group is
trying to decide on the wording of a passage, with nobody taking the initiative to make a
decision. Sarah tries to get the group to contribute, without success, while Ryan is unable to take
control either:
Ryan OK I'll read mine, [reading] if a person is beingavoided and very hurt inside [stops reading], I wroteit twice, I'll fix it now
Sarah I don't get itIlona well ask your group matesCarla it doesn't make senseRyan okay I made a mistake I’ll fix itIlona Carla says it doesn't make senseSarah Guys should I erase this cause it is under Ryan under
a different question, it makes zippo sense where itis under Ryan's question, so should I move it up towhere Carla's is?
group [no response, all are typing]Sarah hello?Sarah fine don't listen to meKate I didn't get that!Carla You want to put the same answer with mine?Sarah all ye of little help? I was asking if I should put
it with Carla's there [points to screen]Carla yakedee yak don't come backSarah OK CarlaKate I don't knowSarah Ryan I'm trying hereRyan oh, forget it, come onIlona why are you having so much trouble today?Ryan I don't know, okay, come on, huddle the computers
together [turns monitor]group uh oh [giggles]
61
[W3 43:48]
Ryan is trying to force the group to work with him, but they don't take these attempts at
controlling the group very seriously. The group has not yet developed an understanding of the
strategies needed to work together effectively.
Different Control Strategies: Silencing vs. DirectingThe shifting of control between Ryan and Sarah is a result of the different strategies used
by Ryan and Sarah to gain and maintain control of the ongoing discourse. Ryan controls the
direction of conversation and discussion by silencing the other group members. He does this in
Aspects by keeping a lock on the text being entered and attempting to direct the order of
contributions from other group members. Looking at the introduction again, it can be seen that
Ryan is quick to cut off any unwelcome contributions from the group:
Carla we're not only covering prejudice we're also coveringminorities
Kate but that's all to do with itRyan but that's all prejudice prejudice links it - we
interviewed different people of different ages onCarla on the topics we have mentionedRyan no no shhh listenCarla Ryan you're gettingRyan sorry [giggle] interviewed people of different ages
on prejudice matters. right?[W11 23:12]
Similarly, when working away from the computers, he tries to keep control of conversation using
a variety of dominating tactics, including standing up, raising his voice, and waving down
conflicting ideas.
Sarah, on the other hand, attempts to control the discourse by encouraging the group
members to make contributions, and by organizing and directing their responses:
Sarah well I’d like to hear what Carla thinks also then wecan all put it together
Carla in this magazine our group has put together ourthoughts on prejudice
Sarah and our information. okay [starts to type] togetherour
[W11 18:39]
When working on Aspects she also uses the built-in control mechanism to maintain her hold on
the text, making little contribution herself. Ryan will often make several attempts to get his ideas
accepted, fail, and the group will stall. At this point, Sarah will step in and take over:
Ryan I think we should have an ending sentenceCarla its not in particular bad to be in minority
62
Ryan you write that [to Carla] you write, you have to goup a line [long pause]
Sarah OK I'll write itRyan No noSarah [gets into position for typing]Ryan being in the minority is not particularly, not
necessarily bad, not always good, some people mightthink that but some may not [long pause]
Sarah OK [starts typing][W3 52:00]
Away from Aspects, she tries to organize both larger tasks, such as who will work on each
assignment, and more specific tasks such as editing and ordering the articles in the magazine.
Both Ryan and Sarah make use of the technology to assist their attempts to control the
group. However, Sarah is more adept at using the parallel social channels to control the group,
following through and organizing once she has control, whereas Ryan falls short, getting
frustrated when he fails to maintain his control. The use of the computers requires additional
skills not necessary in traditional social interaction, but they do not completely remove the need
for social skills. It is the combination of the two which enables Sarah to take control of group
interactions around the computer. The nature of the task, directing and controlling rather than
directly contributing, enables her to avoid her lack of confidence in her own self-generated ideas,
instead controlling the flow of the ideas provided by the rest of the group.
Control In and Around the Shared WorkspaceThe situation with the Tuesday group was quite different. In this group, there were two
strong personalities, David and Faith. Both of them were very confident, both wanted to control
the group, and both would get impatient and frustrated when the direction of the discussion did
not match with their expectations. In social situations, the competition is fairly even; the two
tend to clash, but neither would completely dominate. This could be seen during the
brainstorming session just before the group wrote their story (see Appendix B). Faith is trying to
put forward some story ideas, and David keeps criticizing them, never allowing her to complete
an idea:
Faith you know after the civil war they had to deal withracism towards the black people
David not in all parts of the country, cause every timesomeone considers racism its blacks blacks blacks
Faith I'm talking. there still is, okay, one guy who hatesthem cause he learned that he must hate them. After,this person meets one and he's nice
Elisa and they become friendsDavid why do we always have to do it on a black?Faith I never said anything about blacks, I was talking
63
about the civil war[T5 21:04]
Despite the fact that he blocks Faith's ideas, David never completely controls the group. This is
possibly due to the fact that the other two group members, Sam and Elisa, often side with Faith.
The division of the group along gender lines is uneven, one boy and three girls. The three girls
are grouped together in opposition to David by the very nature of his behaviour, confrontational
and aggressive. However, Sam and Elisa do not always side with Faith. Indeed, more often Sam
and Elisa avoided Faith as well; her position was defined more strongly by her opposition to
David than by her gender.
This pattern of control changed when the group worked on the computer. Here David is
very comfortable, whereas Faith is uninterested in the technology. For example, when writing
the story during the fifth week, Faith dominated the discussion before the group moved to the
computers. She was not entirely in control, as David made numerous contributions. However,
once they moved on to the computers, David took control without any negotiation. There he
assumed the role of scribe, controlling what gets entered in the document. However, unlike
Ryan, he also took an active role in the composition of the story. He followed through on his
control of the discourse, prompting for the next sequence in the story, and adding to it as well:
Elisa okay, one dayElisa its Bagoodas or Bagodas or Bogados or [leaning over
and looking at David's screen]David no we'll put it this way Mrs.Sam Mrs. Bagooda andDavid Mrs.Sam Mrs. MahomedFaith MohammedElisa both had babies at the sameSam both had baby girlsElisa both on the same day as a coincidence [looks over at
David, who is still typing]David okay, by coincidenceSam by coincidence they both came out on the same day
[Elisa and Sam look over at each other]Faith by coincidence on June 25thDavid on a dry day [typing this] not on, in JuneElisa capital J [looks at David's]David it IS capital [points to his]Elisa oh sorryDavid both gaveElisa birthDavid birthElisa to baby girlsDavid to a baby girl
64
[T5 29:14]
David’s use of the role of scribe alters the nature of the role. The scribe, in traditional work
situations, is a passive role. The scribe listens to dictation or a group discussion, and records the
text or discussion without any active input. Even if the scribe is a group member, the written
record is expected to be unaffected by the scribe’s personal opinions. In Aspects, however, the
scribe tends to be regarded as an active, dominant role, with the scribe having direct control over
the content of the text and the direction of the discussion.
David was able to make use of his privileged position as scribe to control the contents of
the document, altering the direction of the text without immediately explaining why. This would
not have been possible in a normal discussion:
David [looks over at others, and says] when women aremarried in India [typing]
Sam [glances at David and back] isn't this off topicthough?
David no, don't worry, we're trying to explain, when they[back to typing]
Sam can't we write this part THEN explain, cause it seemsreally off topic
Faith [not paying attention, looks bored]David no lets do it, just wait, then they pay a what should
we call it, dowry?[T5 30:33]
All of the group members made contributions, although those contributions were subject to
David's approval. This frustrated Faith at times, but she was willing to stay within the protocol
of the situation and work with David as the scribe. Rather than contest him, she withdraws.
David was making deliberate use of the control mechanism in Aspects to maintain
control, even resisting attempts on Ilona's part to intervene and add in paragraph-breaks, which
would have allowed the other group members to make changes:
Sam he's there!Ilona can I suggest you put some more, um [gets up and puts
carriage returns in on David's computer]Faith [gets involved again during the pause, reading] the
Bagoodas did not want to payElisa [starts typing]Sam [does too]David I didn't want to or else they'll start doing funny
things with it[T5 31:15]
In other instances when David did not maintain this tight control, Elisa and Sam would break off
into a subgroup and enter their own text, in a different location, in an attempt to get their ideas
65
acknowledged.
It is interesting to compare this with the way the group interacts when a subset of the
entire group is present. In one of the earlier sessions, David was away. All three girls were
working on a discussion of different perceptions, and were initially working with Sam acting as
scribe. Sam and Elisa were working closely on the content, but Faith, who had lots of ideas and
tended to explain them by using long narrative examples, was not being included equally.
Impatient, she decided to enter her ideas directly. After she began doing this, Elisa did the same,
ending with the group all writing in parallel, but discussing their ideas verbally. This
demonstrates the flexibility of the medium. When Faith was away, however, and David was
working with the two remaining girls, they isolated him, working on their own. He seemed
content with this, working on a separate task. However, he would occasionally fiddle with their
text or disrupt the conversation, as if he needed to assert his control despite the fact that he was
working separately. A similar incident occurred when all four group members were present, and
Faith and David were working independent from Sam and Elisa. David used a function of
Aspects that allows one user to control the view on all other screens to disrupt Sam and Elisa's
work, changing their location in the document:
Sam [Sam’s screen scrolls without her doing anything]Faith, leave it alone
Faith I’m not touching itSam [looks over at David's screen] David!David what, look I know what I'm doing don't worry [David
is controlling the view on Sam and Elisa’s screens]Sam I can't write anything, mine's stuck, did this
[keyboard] come unpluggedIlona its crashed, no its not, save it, give it a name
firstDavid its cause she's viewing with me. Lets see what you
people wasted your time onIlona who are you viewing with?Sam is that why it busted?Ilona yesDavid yeah!
[T11 47:05]
The technology serves as a prop to be used by David as part of his framing of identity within the
group. Even when he is not directly part of the interaction, in this case between Sam and Elisa,
he is able to use his knowledge about the system to exert a certain amount of control.
When working with physical documents, the students were able to gain control over a
paper by grabbing it if necessary. In Aspects there was no way to force a shift in control.
However, the group soon learned to overcome the limitations of the locking mechanism by
66
simply using the other person's computer rather than trying to get control of the document from
within the system.
David [takes Sue's mouse when she's not looking, moves herout of a paragraph]
Sam [looks back, sees David]David one sec. stop, let me work on this part.
[T11 09:35]
Although the system gave the scribe explicit control of the document, the other members of the
group were still able to make significant contributions and provide feedback which affected the
contents of the document. In the above example, where David had control of the document,
there were several occasions where other students tried to make changes. Kate would attempt to
alter a sentence, and failing that would ask David to make the change, using the telepointer to
indicate the change. So, in spite of the control mechanisms provided by the technology, the
entire group was able to influence the document.
The Influence of Technology on ControlThe ability to access and modify the shared text highlights the complex nature of the
computer-supported shared workspace. It does not just allow the possibility for tight control by
one person; if the rest of the group is also aware of the nature of the workspace, they can take
advantage of it to enter ideas in parallel with little or no control by the rest of the group. When
all are entering text, the entire group is on equal footing, regardless of social position or skill.
All ideas are equally prominent, and can be readily seen and manipulated by the group in a more
concrete manner than in an unstructured group discussion away from the computers. However, it
was not always possible for group members to start a new discourse, due to the constraints of the
shared workspace, and due to the social pressure to retain a single focus of discourse resulting
from the dominance of the scribe.
The introduction of collaborative technology into a group alters the nature of
communication in terms of control of the discourse and sharing of ideas and understanding. This
is due to the nature of the medium, the skills needed, and the interaction of the group members
with and around the technology. Since the new channel provided by the computer is not
sufficient to carry the entire discourse, the group members must be able to incorporate several
skill sets, working on a social level while incorporating the new medium, the computer. In
addition, the type of control mechanism implemented by the shared editor affords certain
patterns of control within the group; the fact that Aspects allows one person to control a given
paragraph, which is often the focus of the group’s writing task, gives the scribe control over the
67
task and the ongoing discourse.
The presence of the shared workspace also has an impact on the ways in which attention
is distributed between the group members. The importance of the text in the task of writing
together, and its role as a means of gaining control of the group, change the group’s focus of
attention. I will examine the ways in which this alters the group’s interactions in the next
chapter.
69
Chapter 5
Attention In and Around a Shared Text WorkspaceThe use of a shared text editor influences how group members can attempt to control
discourse when writing together. The shared workspace also has an effect on the ways in which
group members seek and retain attention, the ways in which they can avoid giving attention to
others, and the ways in which the discourse itself is grounded. I will now examine in detail the
composition of one document, the introduction of the magazine (see Appendix B), by each
group, and use conversation analysis to look at the effect of the shared workspace on attention.
The analysis will be framed by Derber's notion of attention as a scarce resource which group
members seek for themselves (Derber, 1983) . I will then extend this analysis to a discussion of
the development of common ground, and how this connects to the design of the shared
workspace provided by a collaborative editor.
The Pursuit of AttentionCharles Derber, in The Pursuit of Attention: Power and Individualism in Everyday Life,
argues that the exchange and distribution of attention forms the basis of social interaction.
Competition is the result of a conflict in the pursuit of attention, whereas cooperation is the result
of freely given attention. He defines focused attention as consisting of visual attention and
cognitive attention. The focus of visual attention is the person looked at, normally the person
speaking, whereas the focus of cognitive attention is the person or object that is the topic of
conversation. The allocation and distribution of attention determines who or what will receive
the most attention over the course of a conversation. Thus, attention can be seen as a limited
resource which group members compete to receive.
Attention can be attracted through attention-gaining initiatives, and kept by attention-
sustaining initiatives. In contrast, attention can be given to others by attention-giving initiatives.
Derber asserts that in North American culture individuals attempt to gain attention, but usually
remain within culturally acceptable bounds. This involves initiating topics that focus attention
on themselves, while at the same time indicating that they are listening to others, through tactics
such as maintaining eye contact or making acknowledgments. One can react to others'
conversational initiatives by use of either support-responses, which sustain the other's position as
focus, or shift-responses, which attempt to shift focus to oneself. Focus may also be shifted by
more passive means, such as minimal use of support-responses, not providing sufficient support,
or avoiding or delaying support at key moments, or deferential use, offering the weakest
support-response consistent with civility.
70
This framework can be applied to the conversations that take place between the students
in the Prejudice Project when they are working with Aspects. In normal face-to-face
conversation, Derber claims that "in all verbal interaction, the common visual focus at any given
moment is occupied by the person speaking." In my study, however, the computers provide an
additional focus of attention, giving the students a way of avoiding giving the focus of visual
attention to the speaker. This is possible because the computer, unlike other inanimate objects
that figure in a conversation, is not passive. The computers are themselves a channel of
communication. The ongoing group discussion has its cognitive focus situated on the text being
composed. Because the text is visible, and manipulable, on all the students' screens, it is natural
for the visual focus to also shift to the text. The shared workspace provides an alternative means
of avoiding giving support without breaking the codes of acceptable behaviour within the group.
This use of the computer as a focus of attention has the effect of shifting the locus of
discourse from the group members to the computer screen, a shift which also affects the way in
which the communication between group members is grounded. The fact that the shared
workspace represents the developing common ground between the participants further reinforces
its role as an alternative focus of attention, and gives it strength as a means of gaining control of
the group’s discourse. The combination of the use of the shared workspace as a focus of visual
and cognitive attention, and its role as a means of grounding the group’s shared understanding of
the task of writing together, makes the control mechanism built into the shared workspace a
powerful factor in the flow of control within the group. The shared workspace is no longer just
an artifact used as part of the task of collaborative writing; instead, it has become an important
channel in the communication between the group members.
To explore these ideas, I will examine several features of the groups' conversations. The
most important feature is the focus of visual and cognitive attention. To see how the computer
affects this, I will discuss the way in which the students direct their gaze during the conversation.
I will also look at how this gaze direction is used in conjunction with support-responses and
shift-responses, to see how the computers allow students both to avoid others' initiatives and to
make different use of their own initiatives.
On the final day of the study, each group worked to write an introduction to their
magazine. Ilona framed the activity by asking them to think about what a magazine introduction
is supposed to say, then encouraging them to write an introduction "as a group". Both groups,
after a brief discussion of the purpose of an introduction, immediately began fighting over who
would type and what the first sentence would be. In both cases, this led to technical problems
with the Aspects software, which had to be resolved by myself and Ben. Once these technical
71
problems had been resolved, the two groups began working in much different ways. The
Tuesday group was very absorbed in the computers, composing ideas directly on the screen,
whereas the Wednesday group tended to discuss their ideas first, come to an agreement, and then
enter the ideas on the screen. This difference in the groups' approaches to collaborative writing
is partly due to the different ways in which the two groups approach control, as discussed
previously. However, a different style of collaboration also emerges from the way in which the
two groups make use of the shared workspace to mediate the distribution of attention between
group members. The Wednesday group uses the computers as an alternative focus of attention,
providing a way of avoiding giving group members the visual focus, whereas the Tuesday group
tends to devote most of their visual attention to the computers, making it the primary focus of
attention.
The Computer as an Alternative Focus of AttentionOnce the technical problems with Aspects had been corrected, the Wednesday group
decided to choose one person to do the typing. After a few false starts, Sarah took over, both
Carla and Ryan deciding they would just dictate. For the next few minutes, the group worked
closely together to compose a short paragraph. After approximately 8 minutes, the group began
to get frustrated with their lack of progress, and turned to the researchers for advice. They
eventually decided, at Ilona's suggestion, to each write a separate next section, and then compare.
For this paper, I am going to focus on the 8 minutes and thirty seconds of discussion and
composition between the end of the technical difficulties [W11 18:00] and the point at which the
group asks for assistance from the researchers [W11 26:30]. The group was sitting in the
following order, from left to right: Kate, Carla, Ryan, and Sarah.
Derber asserts that people tend to want to be the focus of attention, and will use any
means to gain and then sustain that attention. However, he qualifies this by stating that people
will remain within the bounds of social convention when seeking and sustaining attention. They
will also respect others' need for attention, through the use of supportive devices such as eye
contact and verbal acknowledgment. I have already examined how the shared workspace alters
the way in which the Wednesday group members attempt to control the group's discourse.
Similarly, the computers provide an alternative focus of visual attention, giving the group
members both an acceptable means of avoiding attention-giving responses and an additional
means of sustaining attention.
Gaze and eye contact can be used for both attention-gaining and attention-giving
initiatives. Sarah, as has been show, tends to try to control the direction of discourse by
manipulating who is able to gain the floor and whose ideas are accepted. This can be
72
reexamined from the point of view of the granting and manipulation of the focus of visual
attention. Sarah makes the most use of eye contact when speaking (see Table 5.1), usually when
attempting to draw out a comment from Carla or Kate. She does this to include them in the
conversation, but will then retain the focus of attention on herself after getting the contribution
from either Carla or Kate. She also uses this as a technique to control Ryan's contributions,
turning to Kate or Carla when he becomes too vocal.
Sarah Ryan Carla Kate total
looks at speaker when spoken to 10 10 6 5 31
looks at own screen when spoken to 2 9 7 4 22
looks at own screen when spoken to
(while typing or reading)
4 1 1 0 6
looks at speaker's screen whenspoken to
0 1 0 1 2
total spoken to 16 21 14 10 61
looks at others when speaking 15 11 6 4 36
looks at own screen when speaking 2 13 9 6 30
looks at own screen when speaking(while typing or reading)
4 6 2 0 12
looks at other's screen when
speaking
0 1 0 0 1
total speaking 21 31 17 10 79
Table 5.1: The Wednesday group’s focus of visual attention.
Interestingly, Carla often steps in when Sarah actually intended to include Kate. For
example, towards the start of the discussion, when Sarah is still trying to assert her control over
the division of attention amongst the group members, she tries to pull Kate into the ongoing
discussion, which has so far been dominated by Ryan:4
Sarah our group has put together. [looks over to group]Ryan =a pict. no.Sarah topics of, Kate [looks at Kate] what do (you (think?Ryan (the topics
under prejudiceCarla (in this
magazine, our group has put together [looks over atSarah]
4For notation used in conversation analysis transcripts, see Appendix A.
73
Kate topicsRyan [looks at Carla] topics (inSarah (topics that would add
(together as prejudiceRyan ([turns to Sarah] in prejudice, that add, that add up
together (asSarah ([looks at Ryan] that add upRyan =that add up as prejudist.
[W11 18:00]
Sarah tried to turn the focus of the conversation to Kate, but Carla, who along with Ryan is
sitting between Kate and Sarah, stepped in instead. Kate supports Carla's utterance, then Ryan
also supports it, but uses this support to shift attention to himself. By turning to Sarah and
speaking directly to her, he effectively shifts the entire group's visual focus to himself. Both
Ryan and Sarah make significant use of eye contact and gaze direction to control the group's
focus, looking at others when speaking and spoken to more than the rest of the group (see Table
5.1).
Use of the Shared Workspace to Deny Attention
However, it is not always possible to make use of these attention-gaining initiatives when
working around the computers. For example, immediately after the above incident, Sarah tries
again to shift the focus to Kate, but Ryan refuses to follow the direction of visual attention to
Kate, looking instead at his computer screen and referring to the text there as he continues with
his suggestions.
Sarah Kate what do you think? [leans back to address her]Ryan [looking at screen] okay I think it (shouldCarla ([leans forward
to look at Sarah and points at screen] our group has(put together talking about prejudice [looks atscreen]
Ryan (inthis (magazine
Kate (in this magazine our group has (put togetherSarah ([leans forward
and looks at Kate] Carla and Kate what do you think? [W11 18:15]
Despite Sarah's repeated attempts to direct the group's gaze towards Kate, both Carla and Ryan
have managed to resist this by instead focusing their attention on the shared workspace.
The presence of the technology lets group members avoid giving attention; instead of
looking at the speaker, the listener can look at the screen, which allows her to minimize her
support response without explicitly denying attention. This alters the codes used by the group,
changing what is acceptable and what is construed as rude. Carla and Ryan both tended to use
74
the shared workspace as a means of avoiding giving attention, often looking at the screen when
speaking or spoken to (see Table 5.1). Here Ryan does this, criticizing Sarah's suggestion
without looking away from the screen. Sarah tries to use a glance to shift attention to herself, but
Ryan ignores her attempt at eye contact, keeping his focus on the computer :
Ryan [looks at screen] but don't but don't make it endlike we hope you enjoyed this play [laughing] orsomething it sounds like something on
Sarah =how do we end it our magazine is called prejudice inthe 90s [looking at Ryan]
Ryan =no that sounds too and put together a magazinecalled prejudice in the 90s [looking at screen]
Sarah ah fine [turns to computer and starts typing] andinterviewed
[W11 22:00]
Sarah eventually gives up, also shifting her attention to the screen. Thus, the computer can serve
as an alternative focus for the group's attention.
However, at least with the Wednesday group, the use of glances, gestures and other non-
verbal cues to provide support for sustained focus of attention are not completely replaced by the
computer. Ryan and Sarah were able to effectively use their positioning side-by-side and their
focus of attention to keep the discussion focused on only 2 participants:
Ryan [reading it] in this magazine, our group has puttogether our thoughts and ideas on topics that coverprejudice matters (um, we, [looks at Sarah] yep
Sarah ([glance at Ryan]Sarah we alsoRyan =we we have interviewed many different people and
gotten their thoughts (onSarah (many?Ryan =we have interviewed um people and put together a
magazine you shall (beCarla ([looking at her screen] we have
interviewedRyan [turns to Carla, waves hands] shush!
[W11 20:30]
Carla tries to deny Ryan support and shift attention to herself, but Ryan overrides this by
physically silencing her. The blatant shift-response on Ryan's part is more compelling than
Carla's use of minimized support. However, by resorting to such an explicit strategy, Ryan is
bordering on violation of the group's norms of behaviour, leading to resentment from Carla.
This does not imply, however, that all interactions that attempt to work on both a social
and a computer-mediated level are bound to create hostility. There were instances where
75
attention was distributed harmoniously between the group members, resulting in smooth
interaction:
Ryan okay [glances at Sarah]. [reading] (in this magazineour group has put together our thoughts andinformation (on different topics
Sarah (on different (and [looks over to group]
Carla that cover [looks at Ryan]Ryan [looks at Sarah] that cover. p:prejudiceSarah =[points to Ryan] prejudice mattersRyan yeah! [claps hands]Sarah okay [typing “prejudice matters”]Ryan [looking at screen] that sounds serious.
[W11 19:15]
Here, Ryan glances at Sarah and acknowledges her previous contribution, then begins reading.
Sarah starts to speak, misinterpreting his glance as a shift, but she immediately realizes her
mistake, since Ryan does not keep his gaze on her. She waits until towards the end of his
utterance, and then starts to make an addition, supporting it with a glance. Carla makes a brief
interjection, then Ryan confirms his position in the focus of attention by looking at Sarah. When
he stumbles, Sarah completes his idea, and he acknowledges this with an exclamation. She then
moves in to type the sentence, reasserting her control without disrupting the flow. This is one of
the rare instances where the use of the computer as an alternative focus of attention is smoothly
integrated with the more traditional support and shift mechanisms.
Position of the Shared Workspace in Social Interaction
In a normal face-to-face interaction, the person in the focus of attention needs to elicit
support from the other group members, either by providing the opportunity for support-
responses, or more subtly by allowing for eye contact. This does not always work when the
computer is involved. Here, Ryan is reviewing the group's work so far, using a glance to secure
support from Sarah for his position at the centre of the conversation:
Ryan okay, ([glance at Sarah, then looks at screen] so,(in this magazine (our group has put togetherthoughts and information on different topics thatcover prejudice matters. we have intervieweddifferent people
Carla ([looks at screen]Sarah ([sigh] ([looks at Ryan]Carla =[looks at screen] different people of all ages
[W11 21:15]
However, the support-response on Sarah's part did not succeed, because Carla makes use of the
shared workspace to give herself a way of avoiding supporting Ryan. Maintaining his visual
76
focus on the screen, Ryan is unable to ensure that he will remain in the centre of attention by
building on Sarah's support, and Carla is able to use the same strategy to undermine him.
Interestingly, Ryan responds by turning his attention to Carla, but then twists this
apparent support-response into a shift-response:
Ryan =[turns towards Carla] different people of all ages,and (have gotten
Sarah (of different agesRyan of different ages (and we have put together their
thoughtsCarla (and we have gotten their opinionsRyan into one magazine called prejudice in the 90's
[W11 21:20]
The focus of the conversation has shifted from Ryan and Sarah to Ryan and Carla. Ryan, by
supporting Carla's initiative, has retained his position. Sarah is only able to regain control by
directly confronting Carla. When she questions Carla's topic, Carla falters, and Ryan shifts
allegiance back to Sarah:
Carla and got their opinions (on what they thoughtSarah ([looks at Carla] oh don't you
think we should tell them how we came together andstuff?
Carla well, uhRyan =yeah yeah [glance at Sarah]Kate [leans forward and looks at group] we can get to that
afterwardsCarla ((its boringSarah ((okay [leans forward] in the 90's we came to this
[W11 21:25]
Sarah firmly reasserts her position of control by leaning forward, looking directly at the rest of
the group, and again repeating the paragraph so far. The strategies being used to negotiate for
control are constantly moving back and forth between the social level and the computer-
mediated level. Participants need to keep up with these shifts to stay in the conversation.
The use of the shared text as represented on the computer screen alters the position of the
computer screen in the group. The group begins to focus more of their attention on the screen
than on the group, using the screen as a referent to ground their discussion:
Ryan [looking at screen] okay I think it (shouldCarla ([leans forward
to look at Sarah and points at screen] our group has(put together talking about prejudice [looks atscreen]
Ryan (in
77
this magazine[W11 18:15]
Here Carla uses the words on the screen to reinforce and position her comments with respect to
the decisions the group has made so far. This gives them validity and helps to bring them into
the shared understanding of the text that is being developed by the group. This tendency to use
the text on the screen as a centre of attention also allows attention to be passed through the use of
glances at other people's screens:
Ryan we interviewed different people and (got, of [glanceat Sarah's screen]
Sarah (of differentages
[W11 22:30]
The computers are being used as an active object, unlike physical documents, which occupy a
more passive role, acting as the cognitive focus of attention but not actively used as a visual
focus or a means of passing and gaining attention.
The Computer as the Primary Focus of AttentionThe focus of attention in the Wednesday group, although often shifting to the computer
screen, still remained largely centred on the participants. The discussion of the introduction took
the form of negotiating a sentence, making a decision, entering it on the computer, then
reviewing it. I will now look at the way in which the Tuesday group differed in their use of the
shared workspace. The Tuesday group tended to locate their discussion much more firmly in the
shared workspace. In contrast to the Wednesday group, they would often all suggest ideas,
which the scribe (in this case David) would filter and enter into the text on the screen. The
content would then be negotiated, with the scribe controlling what gets changed and what gets
accepted into the text. Rather than an alternative focus, the computers tended to become the
primary focus of the group's attention. This difference in focus can be seen in the ratio of
number of times people speak versus the number of times people are spoken to directly (79:61 in
the Wednesday group vs. 60:34 in the Tuesday group, from Tables 5.1 and 5.2).
At the start of day 5, Elisa and Sam initially took control of the discussion when David
was forced out of the Aspects session due to a technical problem. However, as soon as David
was back in the session, he started trying to take control, deleting large sections of the text. He
continued to interfere with the girls' attempts to type, until Samantha gave in and supported
David's position as scribe:
Faith [typing]Elisa [typing]David you guys we're doing the bottom one okay? [looks at
78
Sam]Sam no we're not we're doing the top oneDavid no we're not [smiles, selecting text]Sam right, fine, delete the top part [looks at Elisa's
screen] its the same except the bottom part is proofread
[T11 13:15]
For the next five and a half minutes, David managed to retain control over the text, and the group
worked closely together on the introduction. After this period, Elisa and Samantha became
frustrated with David's control, and began working on a different document [T11 18:45]. David
and Faith were left out of this subgroup, and eventually moved off to work on other tasks. In this
section I am going to focus on the five minute and thirty second period between when David
took control and the point at which Samantha and Elisa broke off from the main group. The
group was sitting in the following order, from left to right: Faith, Elisa, Samantha, and David.
Unlike the Wednesday group, this group tended to discuss ideas after they had been
entered into Aspects. As people gave out suggestions, David would type them in if he agreed
with them, putting the ideas under his control. The group would then discuss changes, but these
changes could not be implemented until the text was altered. As a result of this, the group was
much more focused on the screens, reading the text and commenting on it rather than looking at
each other. This use of the computer as a primary focus of visual attention had a direct effect on
the way the group members made use of attention-gaining and attention-sustaining initiatives.
David makes the most extensive use of the shared workspace as a device to manipulate
the focus of attention within the group (see Table 5.2). David is often totally focused on the
screen while the others suggest ideas and are also looking at their screens. In fact, all of the
group members rarely look at the speaker when spoken to. David, as scribe, also tends to
continue typing when someone else is speaking, letting his involvement in the task of writing the
text of the introduction override his obligation to give attention to the speaker. Samantha also
makes extensive use of the computers as a focus of attention, but unlike David does not combine
this with the use of traditional attention-gaining and attention-sustaining initiatives.
David Sam Elisa Faith total
looks at speaker when spoken to 1 3 1 3 8
looks at own screen when spoken to 2 6 1 1 10
looks at own screen when spoken to(while typing or reading)
9 0 2 2 13
looks at speaker's screen whenspoken to
1 0 0 2 3
79
total spoken to 13 9 4 8 34
looks at others when speaking 4 0 3 9 16
looks at own screen when speaking 2 7 4 4 17
looks at own screen when speaking
(while typing or reading)
1 9 6 2 18
looks at other's screen whenspeaking
0 5 2 2 9
total speaking 7 21 15 17 60
Table 5.2: The Tuesday group’s focus of visual attention.
Since David is typing, he is able to make minimal use of support-responses, refusing to
look at the group members who are giving out ideas. This makes it difficult for the speaker to
keep the group's attention. Faith, who prefers to look over to get the group's attention, is
especially frustrated by this technique:
Sam [reading] We decided to work on the problem wethought that (by reading this magazine we hope that
Faith (by reading this magazine we hope everyone willrealize that prejudice
David [typing]Sam =that prejudice (hurts everyoneFaith (is a major problem that weElisa I think there's too (many -Faith (<[raises voice and looks over to
rest of group] is a major problem that we need tosolve I'm good at introductions is a major problemthat we need to solve
David [typing] Faith [looks over at her] I already wasthinking of that. [looks back at screen]
[T11 13:45]
David totally dismisses Faith's suggestion. His brief glance is enough to direct his response, but
he immediately reverts his attention to the screen, giving her no chance to respond. This shows a
complex integration of the use of traditional attention gaining and sustaining initiatives, and the
new focus of attention introduced by the shared workspace.
David often makes use of glances to focus a response or a suggestion, but does not give
the floor to the recipient of the utterance, reverting his attention to screen immediately:
Faith [looks at Elisa] we don't tell about the problem yeahwhat's the problem?
Ilona =right it doesn't say what the problem is guys?Elisa [looking at screen] oh yeahFaith cause it erased it remember?
80
David =that prejudice [turns to group] hurts everyone [backto screen]
[T11 16:45]
Unlike Ryan, who has difficulty retaining control after shifting the focus of attention to the
computer screens, David is able to make careful use of nonverbal devices to sustain his position.
Elisa also makes use of this approach. Here, her disapproval of Faith's suggestion is
shown by focusing her gaze on Faith for a moment, then shifting right back to her own screen to
deny Faith any chance of rebuttal. This time David totally ignores Faith's utterance.
David on the problem of prejudiceFaith ((say umDavid (([typing]Faith four children, or four. three sixth graders and one
fifth grader (decided to work onElisa ([looks at Faith] oh come on [back to
screen]David [typing]
[T11 16:45]
As with the Wednesday group, the computers are being used as an alternative focus of attention,
allowing attention to be denied to certain group members by using the tendency of the group to
focus visual attention on the computers as an excuse for minimal use of support-responses.
Use of Shared Workspace to Gain Attention
The shared workspace did not only serve as a means of denying support. As was
discussed briefly in the previous chapter, the computer could also be used as a means of gaining
control of the group's discourse. A few minutes into the discussion of the interview, the group's
discussion began to wander off-topic. As the conversation strayed from the material in the text,
David was no longer able to control the discussion. However, David managed to use this
distraction to reassert his control. While the others were looking away, discussing minorities
within the group, David started deleting sections of the text which he disagreed with. Ilona
noticed, and turned the group's attention back to the screens:
David [starts deleting text]Ilona you guys if you look at your if you look at your
document you will see it has changedDavid [smiles, typing]Sam [points to her screen] we lead into, we lead you
into, [glance at David's screen] you forgot the youFaith [looks at Elisa] Elisa you're a minority cause you're
wearing redDavid [selects text, but it is locked by someone else; he
double clicks, selecting the whole paragraph byaccident] stupid thing
81
Elisa no don't erase that! (oh no!Sam (hehe [glance at David's]David [typing, glance at Sam's]Elisa No! stupid! You erased that!
[T11 18:00]
The group members had different responses to David's attempt to shift the focus of attention to
himself. Elisa directly confronts him, complaining about his actions. He ignores her, able to do
this because he is appearing to stay within his role as scribe. Normally, this overt denial of
attention would be construed as rude, going beyond the bounds of social conventions; however,
the presence of the shared workspace, and its legitimate role as a focus of attention, allows David
to keep his visual attention on the computer at Elisa’s expense. Faith is still distracted, and does
not react. Sam, however, plays along with David's control attempt. Rather than confronting
him, she uses a support-response which allows her some degree of influence over the discourse.
She comments on his changes, giving attention to his screen and acknowledging his control over
the text. This seems to be the most successful strategy to use against David's attention-gaining
initiatives, since it allows the discourse to move forward, and keeps the distribution of attention
collaborative rather than competitive. Rather than fighting with David on a different level, Sam
adapts to David's strategy. Sam makes considerable use of the screen as a focus of attention
when speaking (see Table 5.2).
The difficulty Elisa had when trying to use verbal initiatives and gaze to overcome
control centred in the shared workspace emphasizes this need to work with, not against, the
dominant group member. Elisa wants to alter the sentence under discussion, but David has been
ignoring her, giving all his attention to the task of typing. Sam is focused on her screen, giving
ideas, and reacting to changes in the text as David types them. Elisa tries to glance at David to
get him to acknowledge her requests, but he continues to ignore her attention-gaining initiatives:
Elisa I (need to put something [looking at David]David (tell me something [looking at his own screen]Sam =[looking at her screen] what, by reading this
magazine (we hope that you will realize thatprejudice is a major problem
Elisa (after everySam uh, and that we need to solve, that we oh yeah that
we need to solve (2) andDavid [typing]
[T11 14:30]
All of the negotiation for attention is taking place around the computers. Samantha is able to
gain attention by making direct reference to the text on the screen, editing out loud while David
makes the actual alterations:
82
Sam we hope. <It doesn't really make sense. no I think weshould write [points to screen] um >>[reading] byreading this magazine we hope that you will have
[T11 14:45]
Unlike Elisa, Sam is able to work within David's system of control and get a portion of the
group's attention for herself. She makes the most use of glances at other peoples' screens when
speaking (see Table 5.2), bringing in this additional focus of attention as a means of controlling
the group's attention.
Difficulty of Breaking Scribe’s Dominance
Elisa persists in her attempts to use conventional devices to influence the conversation.
Faith also has trouble making contributions, having to ask for changes to be made, rather than
using her social skills to try to get ideas added to the developing text. She repeats her utterances
several times, receiving little or no support from David:
Sam a lot of (different ideas about itElisa (wait what? can you move your pointer? oh
okay sorrySam change that change that! [using telepointer to point
on screen, glance at David]Faith more understand more about this problem. [looking at
her own screen]Elisa can youSam we need to solve this (problem [points at own screen]Faith (more about this problem
[T11 15:00]
This constant frustration with David's monopolizing of the group's focus of attention
leads Elisa to attempt to break off and create a side conversation with Sam, who is sitting beside
her:
Sam [looking at screen]Elisa [turns to Sam] before we lead you into this magazine
thing why don't we introduce [looks at Ilona]ourselves first? [starts typing]
Sam [looking at Ilona] >can we have pictures of ourselvesthat we can stick in?
Faith period.Ilona whatever we do today (do you have a way of doing
that? I don't think soElisa (lead youDavid [leans forward to look at own screen]Sam I have a camera [turns back to her own screen]
[T11 15:15]
Elisa's shift-response fails because Sam does not offer any support. However, Elisa is able to
83
continue to work on the text, since she is in a different paragraph than David. The group is
becoming unfocused, Samantha moving off into a side conversation with Ilona, and Faith not
clearly focused on either David or Elisa. Since Elisa is able to enter her own ideas in the shared
workspace, David can no longer make exclusive use of the computers to retain control. He tries
to get her attention by addressing her directly, but she keeps her focus of attention on the screen,
denying David a support-response. Faith is also focused on the computer screen, and is
beginning to support Elisa. David has to resort to a distraction; he draws the topic of
conversation away from Elisa by using an explicit shift-response to turn the attention to himself:
Elisa [still typing]Ilona we don't have a way of putting pictures inSam I know how I've got a great idea I'll (draw a picture
of DavidDavid (ElisaElisa =yeah (what [still looking at screen]Faith ([looking at her own screen] of this magazine
(we would like toDavid (you know what I [leans forward and looks at group]
at the at the at the front of the magazine we'regonna say by <<David
[T11 15:30]
Elisa failed at her attempt to create a side conversation because no one gave her sufficient
support. However, later in the session, she does manage to break off from the discussion,
working separately with Sam.
Attention in Collaborative WritingThe above discussion shows how complex the addition of the shared workspace makes
the group's interactions. Even though the Tuesday group's conversation is centred around the
text in Aspects, group members still have to take into consideration both normal face-to-face and
computer-mediated techniques for gaining and sustaining attention. When group members are
able to adapt to the shifting locus of control, as was the case with Samantha and David,
interaction proceeds smoothly.
Negotiating through the computer requires different skills, with which the group is only
just starting to become familiar. In the Tuesday group attention and the discussion of the
emerging text are more focused in the computers than with the Wednesday group. This shift of
the focus of discourse alters the character of group interaction, leading to more conflict. One
person tends to control the text, and consequently the discourse, and the other group members
have to work on several levels to negotiate changes in the text. In the Wednesday group, the
discussion was focused away from computers, so ideas were more accessible to change by all
84
group members. Conflict took place on a social level, so the group was able to use
conversational devices with which they were already familiar.
The introduction of the shared text workspace provided by Aspects alters the nature of
group interaction in conversations focused on the composition of collaborative documents. The
computer comes to act as an alternative focus of attention; this focus can be used to avoid
attention-giving and as a device for gaining attention. As can be seen from the Wednesday
group, attention is distributed through the use of shift-responses and support-responses, and gaze
can be used to both direct and gain attention. The computers play a role in this, providing an
additional focus of visual attention which allows group members to avoid support, but also
makes it hard to use eye contact to retain support. To maintain a smooth flow of attention, group
members need to be able to shift rapidly between social and computer-mediated initiatives.
The discussion of the Tuesday group shows that the focus of attention can be almost
exclusively centred on the computers. Group members can make use of occasional glances and
then focus on the computers to retain the centre of attention. To fight this, others need to work
within the dominant group member's paradigm and attempt to gain control using the same
approaches. The discussion of the emerging text is grounded in the shared workspace provided
by the collaborative editor. In the next chapter I will discuss how this change in the medium
alters the costs and affordances of communication.
85
Chapter 6
Grounding and Shared UnderstandingThe shared workspace acts as an alternative medium for communication within the
group's discussion. This can be considered in terms of how the group members attempt to
ground their discourse. I will now discuss the effects of the new medium of the shared
workspace on the grounding of group communication by making use of Clark and Brennan's
(1991) theory of grounding in communication.
The conversations I have presented in the preceeding two chapters take place around a
specific task, collaborative writing. In any conversation, there must be a basis of common
understanding, a set of codes, beliefs, and assumptions which the group uses to reach consensus
about the meaning of utterances. This is especially true when the purpose of the conversation is
to create a text which reflects the group's shared understanding of a specific topic. Since
attention is often focused on the computers, it follows that the communication may also be
grounded to a certain extent in the shared workspace.
The knowledge upon which the group's shared understanding is based is called the
group’s common ground, and according to the collaborative model of communication (Clark and
Schaefer, 1989) this common ground is developed and maintained through moment-by-moment
collaboration between conversational partners. Grounding takes place through the presentation
and acceptance of components of a conversation.
In normal face-to-face interactions, the common ground on which a conversation is built
is formed by the utterances made by each participant. As contributions are presented and
acknowledged, shared understanding is built up in the minds of the group. When the medium of
conversation is changed, Clark and Brennan argue, the constraints placed on the formation of the
common ground and the costs associated with the process of grounding are altered. Clark and
Brennan assert that individuals tend to use the medium which incurs the least costs and places
the minimum constraints on communication.
In my study, conversation takes place face-to-face; however, the ideas being discussed
are also represented in the shared workspace provided by Aspects. In fact, since the computers
act as a focus of attention the conversation is often grounded almost exclusively in the shared
workspace. I will now explore how the use of the shared text workspace alters grounding of
communication, and discuss the relevant costs and constraints from Clark and Brennan’s
classification.
86
Affordances of a Shared Text Workspace For CommunicationThe shared workspace limits certain aspects of communication, adding constraints to the
grounding of communication. These features characterize the communication that takes place
between the group members when working in Aspects (see Table 6.1). Clark and Brennan use
the term constraints, which implies a negative influence on communication. Since the
introduction of some constraints, such as reviewability, are not necessarily limiting, I prefer the
term affordances, and will use this term in the following discussion.
Affordance Description
copresence A and B share the same physical environment
visibility A and B are visible to one another
reviewability B can review A's messages
revisability A can revise messages from B
Table 6.1: Affordances for grounding in a shared text workspace (Clark and Brennan, 1991) .
The shared document occupies an important position in the groups’ focus of visual and
cognitive attention. This implies that the shared workspace needs to provide some of the
copresence and visibility required for effective communication. This can be translated into
awareness of the location and activities of yourself and collaborators in the text, and the feeling
that the text is shared and acted upon by collaborators.
Even after becoming familiar with the notion of a shared workspace and having worked
on the system for six weeks, the students still had difficulty determining maintaining awareness
of the rest of the group, and awareness of their own position relative to the group in terms of
where they were and what they were doing. The limited affordances for copresence and
visibility in Aspects make it difficult for people to answer simple questions such as “where am I”
and “what am I doing”.
There are several things that the students were asking when they asked "where am I?"
The simplest is location in the shared document. However, it is also important to provide some
feedback as to whether the user is in a conference or working alone, and whether the current
document is private or shared with others. It was common for people to become confused as to
whether their work would be seen by others or not, and whether they would be able to see others'
work. This was especially true when someone stopped working closely with the group, and then
later returned. For example, the Tuesday group thought they had been working together to edit a
document, but had in fact been working in separate copies of the document. This situation
wasn't discovered until Sam happened to glance at Faith's screen and notice that her telepointer
87
was not showing up.
Similarly, in a multi-user conference with multiple documents it isn't always obvious
what you are doing at a given moment, leading the group members to ask “what am I doing?”
This can range from confusion as to whether you are telepointing or not, a simple interface
problem, to more subtle concerns, such as whether you are interfering with someone else's
attempts to edit text. The notion of a shared workspace is a difficult concept with which to
become familiar. Lack of any reminder of where others are makes it easy to forget that there is a
shared workspace:
Group [they start to divide the task up by questions - eachgroup member tells the others what they wrote as ifthe others can't see it]
Ilona Everyone can see the same thing. [goes over andscrolls Sarah’s to show them]
[W4 39:06]
The lack of a sense of copresence raises the questions “where are you?”, “what are you
doing?”, and “who did that?” When working individually, it is easy to lose track of what others
are doing, leading students to ask “what are you doing?” This problem was often overcome by
resorting to physical pointing and glancing at each other's screens:
Sarah I’m getting confused. Ryan, what are we changinghere?
Ryan [makes changes, points to screen to indicate what heis doing]
Carla [watches what Ryan is doing]Sarah [doesn't notice Ryan's gesture] Ryan what are we
doing here?Ryan Its gonna look like a poem.Sarah [sees gesture and looks over] Okay
[W3 23:36]
With synchronous shared access to the document, it is possible to enter text or to delete
someone else's text without that person's knowledge. This can lead to confusion:
David [deleting something]Elisa No no don't erase it DON'T! who erased that?Faith Not me I just got in.Elisa [looks over at Sam] Sam?!?David [looks around, says nothing]
[T6 57:04]
Visibility is reduced due to the lack of indications within the shared workspace as to where
others are, what others are doing, and who is making changes. Tracking of where other people
are and what they are doing can come in many forms. Aspects provides bars alone the side of
the document indicating that someone has control of a region of text. This tells you that
88
someone is there, but not who it is.
Using Clark and Brennan's classification of the affordances for grounding, the addition of
the shared workspace affords reviewability and revisability during communication. The ideas of
individual group members which have been entered into Aspects are no longer intangible
messages which are heard once; instead, they are constantly visible on the screen, and can be
referred to by the group to support or refute later ideas. The text can also be altered; this allows
the text to be manipulated to reflect the changing set of ideas shared by the group. However,
since the text is often controlled by one group member, this allows the ground on which the
discourse is built to also be controlled.
In Aspects, ideas are put out verbally, but are also visible in the shared text space. This
gives the ideas a concrete representation. Once an idea is entered, it is visible to all group
members, and can be manipulated as if it had a physical existence independent from its content
or meaning. The control mechanism determines how ideas are entered into the shared
workspace, and how the ideas are manipulated. Unlike verbal interaction, the shared text space
allows multiple points of control and focus, and all the ideas can be reviewed and commented
upon simultaneously by all group members.
This can be compared with discourse that is focused on physical documents. In this case,
the person who is actually holding the document is often in control of the discussion:
Ryan this repeats [points, then grabs paper]Carla no! [grabs it]Ryan its continuous continuouslyRyan this one needs to read this everyone says the same
thing Sarah [starts to read] listen to this andyou'll understand what I mean
Sarah [reading] perception is good in many waysKate good and bad in many waysSarah hold on there are so many things crossed outRyan [takes doc, takes over reading out loud to the group]
"talked the same, walked the same" see its the same,its repeating
[W8 25:55]
Control over the discourse can be transferred explicitly by turning over, or grabbing, the paper.
It is a physical object, something that conveys an immediate sense of who has it and how it can
be used. This is not the case in Aspects, where it is not immediately clear who has control of the
text, nor is there an explicit mechanism afforded by the object for transfer of control. The group
can circumvent the technology's control devices through the use of other, non-technical, skills:
David how do you spell thought?
89
Sam [tells him]David move so I can change it!Elisa I did!David no don't put it there, excuse me [he violates Elisa’s
personal space, overrides the technology, reachesover and moves her mouse]
Elisa but I wanted it like this [she's highlighted thesection to be changed]
David no, but then I can't edit it wait look I can edit itfrom yours! [he proceeds to change it on herkeyboard] its Mohammed not Mohommed, do you knowanything about the prominent Indian cultures?
[T5 49:39]
However, this use of other resources to gain control can lead to conflict, since it breaks the rules
of conduct developed by the group for working in the shared workspace.
There is only one instance of a given physical document. Its position and orientation all
convey cues as to who has control over it, and how the rest of the group should behave with
respect to both the document and the person holding it. Again, this is not the case with Aspects,
where all have the same view of the document, and can read any part of it at any time.
In a face to face situation, such as when the group is sitting around a table discussing
ideas for a story, ideas are put out into the common space between the group for comment and
discussion. Because the ideas are insubstantial, they can be easily discarded. Certain social
skills, such as persuasion and rhetoric, are required to get your ideas accepted. This is similar to
the process of controlling a group, but is not identical. It is possible for group members to
contribute to the overall discussion without directly controlling the direction of the discourse.
Ideas are often placed in the common consciousness and then picked up later. Ideas are sent out
and detached from the originator, and commented on by others. They do not persist or have any
immediacy without work being done on the part of the group members. This form of interaction
could be supported by a shared editor if it provided alternative shared text spaces, with different
levels of structure (see chapter 11).
The content of a physical document is fixed; it can be marked up and written on, but not
changed. The document on the computer is completely manipulable, as long as one has control
over it. However, the computer document has no visible history, unlike the paper document,
which retains the traces of all who have altered it. There is no permanent sense of history in the
document, so once text is deleted it can no longer be used as context for the discourse:
Faith it looks like I haven't done anything so far. I putone question and Sam erased it and put her question,now I don't have anything
90
[T7 17:21]
Faith feels that she hasn't made any contributions, because her ideas do not explicitly show up in
the file, although her ideas have still had an impact on the overall discussion. (For possible
solutions to this problem, see chapters 10 and 11.)
Unlike the traditional use of computers by a group of students, where all the students
gather around one machine and one person types, all of the group can see and potentially alter
the text that is being composed. On the first day, the Tuesday group worked around one machine
to compose a summary of likes and dislikes [T1 34:24]. Unlike the later sessions, where the text
was composed in the shared workspace provided by Aspects, this document was only
immediately visible to David, who was doing the typing. The other group members made
suggestions, but did not have the entire contents available for review, giving David an advantage
when negotiating which ideas would be kept in the document. Aspects overcomes this limitation
of the location and spatial nature of the document.
The computers were set out in a row, with all four students sitting side by side. This
tended to lead to subgroup formation, with people on the ends getting left out. However, this
was not always the case, since the shared document does not require that people sit side by side
to see the same text. When working on a discussion about stereotypes, sitting in a row at the
computers, Sam and Elisa, although they were sitting at opposite ends, managed to work closely
together, focusing their collaboration on the contents of the document [T4 33:18]. However, the
technology can also lead to marginalization, especially when an individual has trouble accessing
the technology, as often happened to Carla, who tended to get confused and have trouble getting
into the shared document.
Costs of Communication Through a Shared Text WorkspaceThere are a number of costs associated with the new medium of the shared text document
(see Table 6.2). These costs influence the choice of medium during communication.
Cost Description
formulation formulate and reformulate utterances
production producing a message
speaker change changing speakers
start-up starting a new discourse
delay planning and revising before executing
repair repairing a mistake
display presenting an object of the discourse
Table 6.2: Costs of grounding in a shared text workspace (Clark and Brennan, 1991) .
91
The costs of formulation and production reflect the time it takes for an utterance to
actually be entered into Aspects. As long as a group member has control of a paragraph, the cost
of formulation and production are minimal. However, if the person who has an idea and wants
to form an utterance does not have control, and especially if someone else is controlling the
paragraph which is the current focus of the discourse, there is a high cost. In fact, the group
member may be prevented from entering the idea.
This difficulty is connected to the cost of speaker change. In an exclusively face-to-face
conversation, turn-taking takes place through the use of attention-gaining initiatives, whereas in
Aspects there is the addition of the software's paragraph locking mechanism, which allows the
scribe to retain complete control of a text fragment. It is always possible to move elsewhere in
the document and add new text, as Elisa did during the writing of the introduction. However,
this does not guarantee that the new text will be accepted as part of the common ground of the
conversation. The text can be entered, but the group’s focus of attention will not necessarily
shift to this new utterance. The contributor needs to make an effort to get the idea incorporated
into the group’s discourse. This can be seen as the start-up cost of creating a new discourse.
In addition, if a participant wants to form an utterance, but does not want that utterance to
be immediately visible, she can move to a different section of the document and enter the idea.
However, there is no delay between when the idea is entered and when it becomes visible in the
other group members’ documents. Although this lack of delay is useful when working closely, it
also removes one of the advantages of written communication, the ability to polish an utterance
before displaying it. The lack of private space in Aspects makes all utterances potentially subject
to group scrutiny immediately, a disadvantage when it comes to collaborative writing (see the
section on private text spaces in chapter 11).
The immediate display of text on all copies of the workspace, and the ability of anyone to
edit that text once the paragraph is no longer locked, provides a certain level of ability for group
members to repair each others’ utterances. However, due to the paragraph level locking
mechanism, text is only available to be changed once the person entering the text moves to a
different location in the text. This makes is harder for group members to work closely and
correct ideas as they are entered. The perception of who has a claim to ownership of a section of
text, or over the entire document, also influenced the way in which group members repaired
textual utterances. Aspects doesn't provide any explicit indication of who wrote a section of text.
However, the students would often assume, especially in the early parts of the study, that the
92
person who typed a section was the only person who could change it.
Similarly, there was a connection between who typed in a section of text and who got
credit for the ideas contained in the text. The scribe usually provided fewer ideas than the rest of
the group. Despite this, the scribe occasionally took credit for the content of the document. For
example, in the Tuesday group, the person who typed a story, David, claimed the next day that it
was his story:
David I wrote the story.Sam No I did.Elisa I did.Faith we all did.David the one about theFaith I made up Tiger Lily.Ilona I thought everybody wrote it.David I wrote it most cause I typed everything.
[T6 32:27]
Although the system did not provide explicit ownership information, the students tended to
continue to identify text with the person who typed it. This showed up most clearly when the
Wednesday group was editing one of the documents. In this case, the sections of the document
were all written in parallel, with each section easily identifiable as belonging to a different
person. Two of the group members, Kate and Ryan, were suggesting changes, but refused to
make the final alterations until the entire group gave permission
Ryan some of these need changes, like grammar, but Ididn't write them, they have to look at it and see.Can I just make the comments?
Ilona you can do it on paper, right here [gestures toprintout] or just make the corrections on here[points to computer]
Ryan I just wrote that students need to fix things. Ithink that it would hard, since I don't know whatthey meant, so that each person who wrote it shouldfix it, this [picks up story printout] was done bythe group so one person could fix it
[W6 09:30]
When the group was working together on a document that had already been edited, the group
members had no reservations about arbitrarily deleting someone else's text without telling them.
This suggests that at this point the group members had come to regard the text as shared, rather
than just owned by the person who typed it. This is an important step in the process of writing a
shared document. As the text comes to reflect the shared understanding of the entire group, the
ownership of the ideas shifts from the individuals who took part in the interactions that built up
the common ground to the group as a whole. The creation of the text in a shared workspace has
93
an influence on the way in which the group reaches this point of shared ownership and
understanding.
Users may want to explicitly provide information to others about their actions, as a means
of displaying the object of communication and using the shared workspace as a common
referent. To afford this type of nonverbal communication, Aspects provides a simple
telepointing mechanism, allowing each user to gesture with a remote cursor of a user-selected
shape. However, the students often found it easier to use physical pointing and gestures:
Sarah Ryan, can you show me what you’re trying to do?Ryan Take a look here. [points to her screen]
[W3 25:53]
Telepointers were limited when compared with physical gestures because they were both unable
to draw collaborators' attention and lacked sufficient information about the person who was
pointing. All this information is available in a simple hand gesture.
The telepointers also tended to be rather distracting; the students often ended up chasing
each others' pointers around the screen. However, Elisa did learn to use them effectively when
proofreading and editing:
Elisa Meant is spelled M-E-A-N-T [points at her screen]David Where is it?Elisa Its, I'll mark it, there I've marked it. See?
That's where it is [uses telepointer] where my littleannoying thingy is.
[T5 37:28]
Having discovered this function, Elisa explains it to the others:
Elisa Say he spelt birth wrong [points with finger, Samlooks] you go to that [moves telepointer there] andgo like that. [wiggles it]
Sam OHHH.[T5 38:05]
The shared workspace encourages this type of consulting and collaborative learning.
The costs and affordances of grounding communication affect the flow of the discourse,
and interact with the distribution of attention between group members. In a group where the
conversation is grounded in the verbal medium, the shared workspace acts as an addition to, but
not a replacement of, normal face-to-face interaction. However, when the common ground is
located more in the shared workspace the computer comes to play a more significant role in the
flow of attention and control. The effects of the computer on attention and grounding must both
be taken into consideration when examining how the groups in our study go about the process of
94
collaborative writing.
The Impact of Grounding on Collaborative WritingThese differences in the medium used in the grounding of communication alter the way
in which the group interacts, and how the group performs the task of collaborative writing. This
can lead to the use of the shared document as a single channel of discourse, where one person
types and controls the discussion. However, the shared document can also act as an alternative
channel of communication working in parallel with verbal and nonverbal social discourse.
Discourse can also move from one level to another, starting in the text and moving out to the
social level, or vice versa, altering the type of ideas generated, who controls the ideas, and what
devices and strategies are appropriate to maintain control.
This can be seen in the way in which the groups approach writing a segment of text.
There were two patterns in terms of the shifting of the level of the discourse: writing then editing
vs. editing then writing. The distinction is in terms of whether the group revises ideas before or
after entering the ideas in the shared document. The Wednesday group, with their focus more on
the social level, tended to discuss ideas, come to a decision, and then enter the text. On the other
hand, the Tuesday group tended to offer ideas, the scribe would enter the text, and then the group
would discuss the options. This had a number of impacts on the way in which the discussion
was controlled. If ideas are discussed and revised before being entered, the common ground is
social, and control can be negotiated using traditional approaches. However, if the ideas are
entered immediately, the scribe acts as a filter, accepting and rejecting ideas at her discretion.
Further, the ideas that are entered, although visible for review by all group members, are not able
to be altered until the scribe decides to allow access. However, this does allow resolution to be
reached before the text is changed, making the process more orderly than would be the case if
there was uncontrolled access to the text.
The computer serves as an additional medium in the grounding of the group's
communication. The shift of the group's attention to the computers alters the grounding of
conversation, moving from a social to a computer-mediated medium. These two roles interact:
the degree to which the computer is used as a common ground influences its usage in the pursuit
of attention, and the role the computer plays in the distribution of attention determines how much
the shared workspace is used as a means of grounding the conversation. The shared
understanding that develops within the group and is reflected in the text being written is
developed in the common ground built up during the group’s interactions. These factors need to
be taken into consideration when studying group interaction around a shared workspace, and
should also be considered when designing software to support collaborative work.
95
Chapter 7
Approaches to Writing in a Shared WorkspaceFrom these observations, we can begin to develop a characterization of the ways in which
people write together using a synchronous collaborative editor. The approaches used by the
group are significantly influenced by the affordances of the shared workspace.
A Model of Synchronous Collaborative WritingIn more realistic situations than the Prejudice Project, groups do not always work
synchronously; rather, the task is broken into segments using various strategies and the document
evolves over an extended period of time. Posner (1992) characterizes the task of writing
together in terms of roles, activities, document control methods, and writing strategies. We can
draw on these observations to help inform our model of synchronous collaborative writing.
Posner classifies participants’ roles along the following lines: writer, consultant, editor
and reviewer. Although this suggests a rigid division of roles between group members, Posner
cautions that roles tend to change over time. In a synchronous writing situation, where a text is
often developed over the course of a single session, roles tend to shift in a more rapid and fluid
manner. It thus makes more sense to think about the activities performed by each individual.
Posner uses a classification of writing activities that closely resembles the stages model
of writing. In our study, we observed the students brainstorming, planning, choosing a writing
approach, assigning activities, combining ideas, resolving conflicting ideas, entering text,
editing, and reaching consensus about the final text. All of these activities could, and often did,
take place at any time in the process of creating a text. The role taken on by a group member at a
given time was closely related to the type of access the group member had to the shared
workspace containing the developing document. Different activities and roles are afforded by
the type of access available.
The issue of document control is important in asynchronous writing; the way in which the
document is distributed to group members, and the relation with how decisions are made about
document contents, can radically effect the text. Control methods include centralized, relay,
independent, and shared. The control mechanism in Aspects allowed one person to access given
paragraph at a time, which afforded a certain strategy for writing.
The strategies used by collaborators reflect the ways in which the different individuals
contribute to the composition of the document, and the amount of involvement of each in the
creative process of developing the text. Posner uses the following classification: single writer,
96
scribe, separate writers, and joint writing. This classification is similar to the ways in which
students worked together in our study, but at a much coarser granularity.
This taxonomy of collaborative writing provides a starting point for the development of a
model of synchronous collaborative writing. We can begin by looking at the different
approaches used by the students. Sometimes the group will work with their focus of attention
concentrated in the same location, and all group members will enter text. More often, there will
be a common focus of attention, but there will only be one person actually entering the text,
taking on the role of scribe. This tended to be the way the students worked with Aspects.
Another common approach was to work in different locations of the document with a common
task focus. Finally, the students would occasionally split off and work on largely unrelated tasks
with minimal feedback. This constitutes an initial classification of synchronous collaborative
writing (see Table 7.1).
independent writing dividing the task into distinct components and working on
those sections individually, with minimal feedbackbetween group members
parallel writing working on different parts of one document while
remaining aware of what others are doing
scribe writing one person enters the text at the keyboard while the restof the group suggest ideas
joint writing all participants take an active role in the manipulation ofthe emerging document
Table 7.1: Approaches to synchronous collaborative writing.
When assigned a group project during their everyday classroom work, the students
generally use the independent writing approach — dividing the task into distinct components
and working on those sections individually, with minimal feedback between group members.
We started by using this approach to minimize the transition from the students' everyday writing,
since they already had to contend with many changes due to the physical setup of our study.
This early writing also provided us with samples of individual work that we could evaluate and
compare to the students' individual in-class writing and to their group work.
Introducing a shared workspace did not immediately lead to close collaborative work.
The students tended to use the parallel writing approach, working on different parts of one
document and occasionally examining each others' entries. This approach was used by both
groups when writing the poem on day 3. When segments were written in parallel by different
individuals, information was often duplicated. The students also noticed that this parallel writing
97
approach often resulted in a fragmented writing style. Here Sarah discusses how to improve the
poem that the group had written in parallel (see Appendix B):
Sam I think we should write the whole thing over becauseit sounds like 3 poems stuck together.
Faith Yeah, I think one person has to rewrite it.Elisa Yeah.
[T3 29:22]
Parallel writing involves monitoring the activities of others, and coordinating content. Control
can be decentralized, or can focus on one individual who coordinates the activities. This type of
work suits the type of control used by Sarah, organizing and directing without any explicit
contribution of ideas.
The scribe writing approach, where one person enters the text at the keyboard while the
rest of the group suggest ideas, was also familiar to the students. This approach is often used by
individuals working simultaneously on one document using traditional writing software. By
making the group document available to different writers on several different computers, Aspects
subtly alters the dynamics of the scribe approach. Even though a document was accessible on
several computers, in the early stages of the project the group would still gather around one
machine with the scribe controlling the document and making all the final decisions. This is the
way the scribe approach is normally used with a single-user word processor. For the majority of
the writing tasks, both groups used the scribe writing approach.
However, Aspects makes the shared document visible on all users' computers, letting all
the participants have an equal opportunity to view the text. After a few weeks, when the students
had become used to the idea of a shared document, they would follow the text being typed by the
scribe on their own computers, suggest changes to the scribe, and use telepointers to indicate
where the changes should be made. Despite the fact that the scribe still had control of the
document, other group members provided input that led to very fine grain changes in the text,
including spelling, wording, and grammar. The successful use of this form of close collaboration
depended on the ability of the group members to combine the access control provided by the
system, the role of the shared workspace in the group’s interactions, and the social skills needed
to negotiate an understanding of the changes being made to the text.
Later in the project, individuals were seen correcting the text which was just typed by the
scribe as soon as it was accessible to them. This is the joint writing approach. Joint writing
involves all participants taking an active role in the manipulation of the emerging document, a
role not possible in traditional single-user word processors. Joint writing was difficult to
accomplish in Aspects, given the fact that only one person could be actively making changes to a
98
specific paragraph at one time; however, with short paragraphs it was possible to be working
very closely, with all changes visible on the same screen. The students occasionally shifted
control to allow others to make changes, but this was a rare occurrence. Joint writing is also hard
to accomplish in a situation where all group members are not give equal control of the discourse.
In both the Tuesday and the Wednesday groups there was a constant shifting of control, with
individuals trying to dominate and others resisting. This type of interaction does not afford joint
writing.
several people writing
one person writing
same text
different text
scribe
paralleljoint
independent
Figure 7.1: The space of collaborative writing approaches: text and writers
99
different discourse focus
same discourse focus
same text
different text
scribe parallel
joint
independent
Figure 7.2: The space of collaborative writing approaches: text and focus
The choice of writing approach is important, and difficult. Traditional writing tools do
not afford the use of the joint writing approach. The scribe approach gives one person control
over the discourse, and thus leads to a specific style of group interaction. Parallel writing is also
a new form of writing introduced by the shared workspace, but is not as unfamiliar as joint, since
it closely relates to independent writing. These approaches should be viewed as four points in
the space of collaborative writing possibilities (see Figures 7.1 and 7.2).
Real collaboration may include a combination of several approaches (see Table 7.2), as
can be seen by examining the ways in which the two groups approached writing three documents
(see Appendix B). For example, while writing the story, both groups used primarily the scribe
writing approach, but there were some instances of joint writing. As they gained experience with
the technology and the group writing approaches, the students in the first group expanded their
repertoire. While writing the magazine's introduction on the eleventh day, each member of the
group worked inside the shared document, writing quietly and making changes to his/her entries
based on each others contributions but without much discussion. At times students would glance
at other parts of the document or discuss ideas, assuming that everyone was aware of the entire
document's contents. Editing took place during the writing, as opposed to as a distinct stage.
Writing
Samples
Wk
# Plan
Writing ApproachWrite Edit
GROUP poem 3 — parallel [5 minutes] joint [7]
100
T story 5 joint
[6]
scribe/joint [19] scribe/joint [10]
introduction 11 joint[1]
parallel/joint [23] while writing
GROUPW
poem 3 — parallel [7] parallel/joint [9],joint [6]
story 5 joint
[5]
scribe/joint [20] joint [11],
independent [2],joint [3]
introduction 11 joint
[3]
scribe/joint [11],
parallel [3], scribe/joint[11]
while writing
Table 7.2: Writing approaches at different stages in the composition of 3 documents.
As they became more experienced with collaborative writing, the students also became
more selective in their use of writing approaches, to the point of switching strategies in the
middle of composing a document. For example, when working on the magazine's introduction,
the Wednesday group shifted between scribe/joint and parallel writing to solve a consensus
problem. During the scribe/joint writing component, the group couldn't agree on the best
wording for one section of the document. To solve this, three members worked in parallel to
enter their own ideas, while Sarah directed their activities. Once the ideas had been entered, they
all read these entries and combined them into a single paragraph. The group then shifted back to
scribe/joint writing to complete the document.
Collaborative Writing and the Models of Individual WritingThe models of individual writing can help to inform our understanding of collaborative
writing. The stages model of individual writing breaks writing down into three stages: plan,
write, and revise. The text is seen as a preexisting object which the writer translates into words.
Clearly this does not reflect the way in which writing is actually performed, but it does highlight
the important steps which are iterated through during writing. Planning is constantly being
carried out as the text is formed, both in the minds of the writers and on the page or computer
screen. In a collaborative writing situation, planning occurs internally, and is externalized
through spoken utterances and tentative contributions to the physical or electronic text. Thus, the
planning stage, in collaborative writing, starts to merge with the writing stage, as plans and ideas
are made concrete in the developing text. As others see these ideas being entered into the shared
workspace, they compare them with their own ideas and plans, and attempt to resolve the
resulting conflicts through a revision of the plans and corresponding text.
101
This is similar to the problem-solving model of individual writing. According to Flower
and Hayes (1984) , writing can be viewed from an information-processing perspective, where
writers have a set of goals, and methods which can be used to achieve those goals. As the text is
formed through an iterative process of planning, writing and revising, the writer tests the
developing text against the established goals, monitoring for inconsistencies; when a problem is
found, the inconsistency is diagnosed, and then methods are fired to correct the problem and
bring the text back in line with the objectives. This model incorporates the uncertainty involved
in writing together, but does not consider the social context of the writing process, specifically
failing to account for the role of the reader.
The process of writing, and especially writing together, can be seen as analogous to a
conversation. As described above, the iterative process of putting forth an idea or plan, having a
collaborator attempt to understand that idea, and then revising the idea based on the group’s
reactions, is very similar to the process of making an utterance, seeing a conversational partner
respond to that utterance, and then reinforcing or correcting the utterance to retain shared
understanding between the participants in a conversation. From this perspective, the social-
interaction model of writing most closely models the process of writing together. According to
Nystrand (1989) , writing involves not just the author, but the reader, and is a process of
attempting to achieve a shared social reality represented in the medium of the developing text.
When writing, the author initiates a written discourse, and then tests for reciprocity. If there is a
breakdown in shared understanding, the author elaborates or rephrases the textual utterance in an
attempt to move back onto the developing common ground.
In individual writing, this process takes place between the author and the anticipated
audience. However, in collaborative writing the roles of author and reader become blurred.
Each author has an audience in the form of the other authors, who are able to evaluate the
developing text from the position of the reader as well as the writer. Authors need to
communicate about the topic and attempt to reach a common understanding of the content of the
text. Then the group needs to transform this understanding into text. To do this, the group needs
to be able to manipulate and view the developing text, and communicate about both form and
content. This involves referring to the text through gestures and physical marks on the text, as
well as verbal references. Throughout the course of the process of transforming the group’s
ideas into a textual form, the group needs to reach a consensus about the content and form of the
text, which will eventually come to reflect the shared understanding of the group. This process
involves a negotiation for control, attention, and meaning within the group. Collaborative
writing makes writing an explicitly social process.
102
Part III: The Design of a Collaborative Writing Tool
In the previous section I presented the results of the Prejudice Project, discussing how the
introduction of a synchronous collaborative editor into the social fabric of groups working
together to write a text document alters the nature of communication. I looked at the impact of
the shared editor on control, distribution of attention, and grounding of discourse, and
characterized the writing approaches used by the groups when using the Aspects shared editor.
These results can be used to inform the design of new computer-based tools to support
synchronous collaborative writing. In this section, I derive requirements for a collaborative
editor from my observations. These requirements are grouped into three categories: concurrency
control, awareness, and shared task spaces. The design requirements are explored through the
implementation of Calliope, a prototype shared editor, which I describe in detail.
103
Chapter 8
Implications for DesignDhar and Olson (1989) argue that most systems that attempt to support collaboration
have a limited view of that support, addressing either communication or problem solving, but not
both. The emphasis tends to be on the provision of a shared object, not support for
communication (Leland, Fish and Kraut, 1988) . This is particularly true for collaborative
writing systems, which tend to emphasize task at the expense of communication. As Derycke
(1992) suggests, in collaborative systems, the synchronous document is the object of
communication. A synchronous shared editor introduces the idea of a shared workspace, a
location in which writers can ground their ideas and develop a shared understanding of what they
are trying to convey. The design of a collaborative editor needs to support both task and
communication requirements.
Requirements for a Synchronous Collaborative EditorI have described how groups work together to create a text in a synchronous, face-to-face
situation. I have also examined the impact of the use of a computer-based collaborative writing
tool on the communication that takes place during collaborative writing. From my analysis of
the Prejudice Project, I have shown that there are 3 categories under which the collaborative
writing process can be organized: the way in which control of discourse is negotiated amongst
the group members; the role of the shared workspace in the group’s focus of attention; and the
techniques used by the group for grounding of communication during their discourse about the
task of writing together. These factors all influence the writing approaches used by the group,
and can be used to inform the design of a synchronous shared text editor (see Figure 8.1).
control of discourse
focus of attention
grounding ofcommunication
concurrency control
awareness
shared taskspaces
Figure 8.1: Impacts of shared workspace and resulting design requirements.
To provide computer support for synchronous collaborative writing, there are a number
of requirements that the design of a collaborative editor must satisfy. Important issues in the
104
design of a synchronous shared editor include support for a shared artifact, which includes
decisions about how to deal with concurrency control mechanisms and how these mechanisms
affect interaction; awareness of the activities and location of collaborators in and around the
shared workspace; and communication in and around the shared task spaces provided by the
system. The requirements for dealing with these issues emerge from the observations of how
people write together in a shared text workspace.
Concurrency Control
Traditionally, control in collaborative writing takes place between group members
through the use of rhetoric, conversational devices, and attention-gaining and attention-giving
initiatives. The addition of the shared workspace provides an additional means for controlling
the discourse. To control the discourse, an individual needs to control both the attention of the
group and the common ground on which the discourse is built, in this case the shared workspace.
The access control mechanism in the shared workspace can afford specific types of
control. In Aspects, the paragraph-level locking mechanism encourages the use of the scribe
writing approach, with one person entering text, and consequently controlling the ideas that get
represented in the shared workspace. The way in which the scribe controls the group varies from
the use of silencing to block others’ contributions, to the directing of the group by eliciting ideas
and then choosing which are accepted. These were the forms of control used by Sarah and Ryan
in the Wednesday group. David used a combination of the two, silencing the group through the
use of the shared workspace, and directing the discourse by manipulation of the group's
attention.
Requirement 1: provide flexible concurrency control mechanisms
A collaborative writing system provides a shared text workspace, and needs to maintain
consistency across all users’ views of the shared workspace during concurrent access to the text.
The choice of control mechanism directly affects the approaches taken by a group to writing
together, and influences the ways in which group members can control the group’s discourse.
This implies that the system must provide flexible concurrency control mechanisms. This
requirement addresses the basic need for a shared context, and the need for flexibility in terms of
how the shared context is maintained and conveyed to the members of the group. By providing
flexibility in terms of concurrency control, the group rather than the system may determine the
style of collaboration.
Awareness
When working together using a collaborative editor such as Aspects, the group’s focus of
cognitive attention is the developing content of the text. The workspace can act as an alternative
105
focus of attention, serving as a way of avoiding giving support without breaking the codes of
behaviour and seeming rude. This was the behaviour which was exhibited by the Wednesday
group. It can also come to act as a primary focus, depending on how much the group tends to
ground their work in the shared workspace, as was the case with the Tuesday group. This allows
the workspace to be used as a means of gaining attention.
Requirement 2: enhance peripheral awareness of presence and actions of group
The fact that the shared workspace is very much a part of the visual focus of the group’s
attention alters the degree of communication that takes place during writing. With attention
focused on the text, there is less awareness of the conversational partners. Retaining attention on
the text during a discussion makes it difficult for participants to monitor for the actions, attitudes
and non-verbal signals of the rest of the group.
The use of the shared workspace as a focus of attention suggests the important role of the
workspace in the grounding of the group’s discourse during the process of writing together. The
use of the workspace to ground the group’s discourse introduces a number of affordances for
communication through the text, including copresence and visibility of collaborators. A
collaborative writing system needs to address the problem of awareness in the shared workspace
to maximize the affordances for communication. The system should enhance peripheral
awareness of the presence and actions of the group: users should be able to maintain a
continuous, implicit awareness of where others are, what others are doing, and how others are
affected by their actions.
Requirement 3: maintain a sense of the history and context of the developing text
The shared workspace allows the entire group to see and review the developing text.
This makes the ideas more visible, and allows them to be referred to at any time. The concrete
representation of the text also makes it theoretically accessible to all for revision; however, the
mechanisms used by the collaborative editor to provide access control can influence who
actually gets to alter the text. If text is altered, there is no history of the context of the
conversation readily available. The ideas actually visible in the shared workspace are more
important than those no longer in the text, as opposed to verbal discourse, where all previous
ideas have equal status. The affordances for reviewability and revisability affect the
communication which takes place during writing.
The communication which takes place around the text as the group is writing is sustained
through reference to the text. When they were writing their magazine, the students in the
Prejudice Project wanted to refer to previous versions of the text to justify their suggestions and
106
revisions, and to keep track of individual contributions. They made assumptions about who
wrote the text, and changed their view of ownership of the document as the text was revised.
This suggests that the system should maintain a sense of the history and context of the
developing text, to enable users to request information about the text and the actions performed
on the text, and to help the group develop a shared understanding of the development of the text
in its current state. This awareness and context is essential to support the ongoing discourse both
around and through the text.
Shared Task Spaces
When writing together, a group is working to develop a shared understanding of the
content and form of the text to be written. The presence of a shared document, visible on all of
the group members’ computer screens, provides a concrete representation of the text as it is
developing. This changes the way in which the group develops the ideas, and the way in which
these ideas are transformed from concepts into text. The system should provide shared task
spaces which take into consideration the role of the shared workspace in communication.
Requirement 4: reduce the costs of grounding
The shared workspace, as a powerful focus of the group’s attention, and a concrete
representation of the developing text, acts as part of the common ground on which the group’s
shared understanding is built. However, as I have shown, there are a number of costs associated
with the use of a computer-based medium for grounding.
The costs associated with speaker change, start-up, and formulation, production and
repair of utterances all result from the fact that the shared workspace affords certain forms of
access to and control of the text. The representation of the ideas in the text may or may not
actually coincide with the group’s understanding of the material, especially if a subset of the
group, or more likely one individual, has control of the manipulation of the text. This
discontinuity can lead to conflict and the breakdown of collaboration. The workspace also
introduces the cost of display, and the cost of lack of delay in the transmission of utterances.
Difficulty in displaying ideas reduces the use of referentiality, as seen in the Prejudice Project
when students had to lean over and point at each others’ screens to make reference to ideas. The
lack of delay in transmission makes it difficult to polish text before presenting it to the group.
These costs all reduce the effectiveness of communication during collaborative writing.
The shared task spaces provided by a shared editor should address the problems of
communication, control and grounding introduced by the presence of a shared workspace, in an
attempt to reduce the costs of grounding in the new medium of communication.
107
Requirement 5: provide multiple related channels of communication
The shared text, with its role as a representation of the group’s ideas and understanding,
provides a channel for indirect communication between the group members. Ideas entered in the
shared workspace as a discussion proceeds, the use of the text as a referent, and the utilization of
the document’s structure and form to organize the group’s ideas all provide reinforcement for the
direct communication which takes place around the text. This means that the group’s discussion
takes place not only around, but also through, the text. During the Prejudice Project students
were able to do this by discussing ideas, entering the text in the shared workspace, and then
manipulate and refer to the text as they formed their shared understanding of the content of the
text.
The single, structured text space with paragraph-level locking provided by Aspects
limited the ways in which the groups could communicate through the text. A shared editor
should include various shared spaces in which to work, rather than just a single shared text
document, to provide multiple related channels of communication. This will allow the system to
accommodate different writing approaches and the need for group members to work at different
levels of task and discourse focus throughout the process of writing together. It will also allow
the group members to use several channels for communication when presenting an idea,
providing reinforcement and encouraging the use of referential communication.
From Requirements to Implementation...
These design requirements provide us with a framework in which to implement a
prototype collaborative editor which attempts to meet the requirements of communication in the
process of writing together (see Table 8.1). The challenge is to develop a system which not only
provides the basic resource sharing necessary for collaborative writing, but also includes the
appropriate levels of sharing and access control, supports awareness of the presence and actions
of the group, and provides the means with which to communicate ideas and ground those ideas in
the shared workspace.
concurrency control • provide flexible concurrency control mechanisms
awareness • enhance peripheral awareness of presence and actions of group
• maintain the history and context of the developing text
shared task spaces • reduce the costs of grounding• provide multiple related channels of communication
Table 8.1: Summary of design requirements
Calliope: a Prototype Shared Editor
108
To explore the implications of the design requirements laid out in this chapter, I built a
prototype shared editor. Calliope5 provides a main shared text workspace, a number of tools to
enhance collaborator awareness, and a variety of alternative shared workspaces designed to
afford communication and reduce the costs of grounding. I will now describe Calliope, and
discuss the issues and design decisions involved in the development of this system.
Implementation Strategy
Calliope was built to explore areas such as resource sharing, awareness, and
communication, and the related human interface issues. Although technical issues were part of
the investigation, my focus was not on the implementation issues involved in the development of
a shared editor from scratch. This has been done a number of times before. At the University of
Toronto, the SASE (Mawby, 1991) and SASSE (Nastos, 1992) prototype collaborative editors
had been built, demonstrating the difficulty and complexity of the technical problems, exploring
some possible solutions, and investigating the interface design problems. I wanted to be able to
build a minimal shared text editor, and then use that system as a basis for experimentation with
the above design requirements.
Therefore, I needed an implementation framework which would both provide the basic
network and distributed computing support required for groupware, and at the same time allow
the development of rapid prototypes with complex graphical interfaces. The predecessor to
Calliope, SASSE, was built on top of a single-user text editor written at the University of
Toronto. This was combined with a socket library, also originally written at Toronto, to add
network functionality. The rest of the system was built in MPW C (Apple Computer, 1988b) on
top of the Macintosh Toolbox (Apple Computer, 1988a) . This system was a successful
prototype, and provide the inspiration and much of the background for Calliope, but was not in
any shape to be used as the starting point for the new system.
For a number of reasons, I initially wanted to continue to develop on the Macintosh
platform. To begin with, the user interface is much more consistent and integrated than any
other platform. The target user base for the system, the students at the Huron School, use
Macintoshes, and we intended to interface the system with CSILE (Scardamalia and Bereiter,
1992) . Therefore, I began an implementation on the Macintosh. Development tools had greatly
progressed since SASSE was built; I used the MetroWerks CodeWarrior C++ (MetroWerks,
1996) package, building on their PowerPlant class library and application framework. Network
5In ancient Greek mythology, Calliope (Cal-EYE-oh-pay) was a MUSE (Multi-User Shared Editor), one of the nine
daughters of Zeus and Mnemosyne who provide inspiration to writers and artists. Calliope was the patron of writersof epic poetry, the predecessor of modern fiction, and would no doubt have encouraged people to write together.
109
support was provided by the GUSI socket library (Neeri, 1996) , a distant descendent of the
socket library used in SASSE.
It soon became obvious that the overhead involved in developing a complex system on
the Macintosh was not conducive to the rapid prototyping and experimentation needed to
develop a new shared text editor. The implementation got to the point of a simple shared text
workspace, providing serialized concurrent access through a centralized communications server,
with coloured remote selections for collaborator awareness (see Figure 8.2).
Figure 8.2: Initial version of Calliope.
The effort required to get to this point did not bode well for further experimentation. As
an alternative, I began to consider GroupKit, a groupware toolkit developed at the University of
Calgary (Roseman, 1993; Roseman and Greenberg, 1996) . GroupKit provides the basic support
needed for a groupware system: connections to remote processes, session management, a number
of abstractions to support distributed processes, and a selection of interface widgets to provide
some of the functionality required in a shared application. I had initially considered GroupKit,
but decided against it because it runs on top of tcl/tk (Osterhout, 1994) , an X-Windows based
toolkit that was not, at that time, available for the Macintosh. However, since a port of tcl/tk to
the Macintosh and Microsoft Windows was in progress, and GroupKit would provide a much
more robust and flexible environment than PowerPlant and GUSI, I decided to shift my
implementation strategy.
System Architecture
GroupKit provides a number of mechanisms to support the development of collaborative
applications, including session management, inter-application communication, and a simple
application framework. GroupKit applications consist of a number of replicated processes,
110
called conferences; a collection of conferences that represent people working together on a single
task are called a session. Session management is coordinated through a registrar, a central
process which provides a connection point for new users, and maintains a list of all participants’
identities and locations. The registrar also optionally maintains the content of sessions after all
users have left, allowing for a certain amount of persistence between active sessions.
Each user runs a local process called a registrar client which allows the user to create,
delete, join or leave conferences, and to see a list of active sessions and their participants. The
registrar client makes requests on behalf of the conference to the registrar, and creates new
instances of conferences. The basic registrar client provided with GroupKit to handle session
management, called the open registration client, allows any user to create a session, and any user
to join an existing session at any time (see Figure 8.3). It also allows all users to see all currently
active sessions. Since I am not focusing on the initiation and joining of shared work session, I
elected to use the open registration mechanism in Calliope.
Figure 8.3: Session Control using open registration.
The conference is the shared application built using GroupKit, and is separate from the
registration system. Once created, conferences maintain direct connections to each other, which
are used exclusively for communication during a session. GroupKit provides primitive
operations to support point-to-point and multicast communication between conferences. The
central registrar is only used for session management, and plays no part in inter-conference
communication.
111
This framework provides the necessary functionality for the implementation of simple
shared applications. The current implementation of Calliope is built on top of this framework,
running on a SUN workstation under X-Windows, and in a minimal fashion on the Macintosh.
GroupEdit attempts to address the design requirements derived from my observations of
student writers during the three months of the Prejudice Project. The system provides a main
shared text workspace, with a variety of access control mechanisms, ranging from selection level
to paragraph level locking. It provides a number of awareness tools, including shared scrollbars,
a fisheye-lens gestalt view, and text colour-coded based on who entered the text. It also provides
a number of different windows into the main text, and a variety of parallel shared task spaces,
including private text windows, an unstructured text scratchpad, metatextual annotations, and
external annotations. Finally, since GroupEdit runs within the GroupKit environment, it can
easily be combined with other groupware tools, such as brainstorming tools, chat windows,
shared drawing and sketching tools, voting and post-it note tools, and concept maps. This
provides a powerful range of tools for a group to use during the process of collaborative writing.
I will now discuss in detail the issues involved in satisfying the three areas important to
the design of a synchronous shared editor, and describe how Calliope was used to explore the
implementation of these requirements in a prototype shared editor.
113
Chapter 9
Concurrency ControlCollaborative writing involves several people working together, possibly at different
times and in different places, on the same text. The degree to which authors share the text being
written can vary greatly depending on the way in which the group approaches the task of writing
together. In the case study I have been discussing, the authors were working at the same time in
a shared workspace. The students usually worked in the same location, but occasionally worked
on different parts of a document, either together or separately. All of these situations involve
several people making changes to a shared document. Some thought must be given to how the
different users' views of the shared text are kept consistent, and how the contents of the shared
document are kept from coming into conflict during concurrent access.
Concurrency control problems arise when a computer program’s software, data and
interface are distributed over several computers. The fact that users can perform operations on
shared objects which may be executed in different orders on different computers introduces the
probability of conflict. This conflict would destroy the commonalty of the shared workspace,
leading to confusion in the group, and a breakdown of the collaboration. For groupware to
effectively support cooperative work, this problem must be avoided.
Ellis and Gibbs (1989) suggest responsiveness and notification as the two human
interface requirements to consider when dealing with concurrency in groupware.
Responsiveness refers to the immediacy of local system reaction to local user actions, whereas
notification refers to the immediacy of remote propagation of local user actions. In interactive
systems, responsiveness to local actions should be free from delay. When dealing with
synchronous systems, users should generally be notified of other users' actions with as little
delay as possible.
However, as we have seen, when involved in collaborative writing users do not always
work in close proximity all the time. The approach taken to writing together determines the
amount of concurrent access to the text. A collaborative writing system should be flexible
enough to allow for any and all types of synchronous collaborative writing: independent,
parallel, scribe or joint. This suggests that the required granularity of control and amount of
notification can vary throughout the process of writing together. I will now look at the different
approaches that can be taken to support concurrency control in synchronous collaborative writing
systems, and discuss how these approaches afford different writing styles.
114
Approaches to Concurrency ControlIn a distributed computer system, processes on separate machines may be attempting to
access shared resources concurrently. In this situation it is important to ensure that the results of
concurrent access are consistent on all machines. When actions on shared resources are
potentially conflicting, some form of synchronization must be performed. There are several
ways in which distributed systems attempt to do this. If concurrency is not critical, then mutual
exclusion will ensure data integrity. If concurrency is to be encouraged, then more complex
forms of concurrency control, involving atomic transactions and either serialization or locking,
must be used. Greenberg and Marwood (1994) present a detailed account of the human
interface tradeoffs involved in the use of different approaches to concurrency control in real time
groupware. Their discussion of concurrency control techniques formed the starting point for the
following discussion. The references to distributed systems techniques for concurrency control
are well-known, and in this case are drawn from Silberschatz and Galvin (1994) .
No Concurrency Control
When interaction is structured on a social level, the collaborative style of the task may
take care of concurrency control external to the application. Often turn taking is enforced by the
style of cooperation, such as in a mediated debate or formal meeting. In the rare situations where
turn taking is violated, conflicts can be resolved by negotiation and manual intervention. Some
systems, such as Colab (Stefik, Foster, Bobrow, Kahn, Lanning and Suchman, 1987) , attempt to
identify these rare occasions and alert the group to the possibility of inconsistencies. This
imposes a certain amount of overhead on the group, requiring that they be aware of the need to
resolve conflicts, a concern which may hinder writing.
Lack of concurrency control is also acceptable in communication-focused applications.
For example, in a brainstorming tool, it is getting the ideas down fast that matters; slight
inconsistencies will not by noticed by the users. The constraints placed on interaction by
concurrency control are more likely to hinder the free flow of ideas, and should be avoided. An
unstructured text space, which affords the rapid entering of ideas without imposing any
document structure, could make use of this approach.
Finally, concurrency control can safely be ignored in systems where the level of
interaction and the granularity of changes to objects is such that inconsistencies are not noticed.
As an example, GroupSketch (Greenberg and Bohnet, 1991) provides no concurrency control.
Small-grained inconsistencies (at the pixel level) can occur, but users reportedly did not notice.
This approach does not work when noticeable differences can occur, such as in a text editor.
115
Mutual Exclusion
The simplest way to enforce control of concurrent access is to use mutual exclusion of
critical sections of program code. Although this does not generalize well to distributed
environments, since it greatly reduces parallelism and concurrency, it does ensure controlled
access to shared resources. Control can be provided by a centralized coordinator, through
consensus in a fully distributed environment, or by use of a token-passing mechanism. Since it
reduces concurrency, mutual exclusion is not a feasible approach to use for a synchronous
collaborative editor, except when the document is highly structured such as in an outline view.
Atomic Transactions
Concurrent access to a shared object requires that, at a certain level, actions on the object
not be interleaved. When execution within a critical section must be performed atomically, the
notion of atomic transactions is introduced. A transaction is a set of operations which are either
all completed and committed at all sites, or are aborted, meaning none of the operations are
carried out at any sites.
In a situation involving concurrency, such as in a distributed system, atomic actions may
overlap. To ensure atomicity, these overlapping transactions must be serialized, meaning that the
concurrent execution of transactions must be equivalent to the case where these transactions are
executed in some imposed sequential order. Mutual exclusion limits concurrency, so while it
solves this problem it does not let us take advantage of the distributed nature of the system.
Mutual exclusion and atomic transactions are very much in the spirit of traditional
distributed systems, where the fact that resources are being shared is hidden from users. This is
the opposite of the intention of groupware such as a shared editor. In a situation where all users
can see the objects being shared, it is difficult to use constructs such as aborted operations, since
the interface level effects can be very confusing. These methods can be adapted for use in a
shared editor, but care must be taken to consider the effects on the interface, and on the social
interactions in the group during writing.
Serialization
Serializability can be maintained either by synchronizing event interleaving so that
atomic transactions are executed serially across the entire system, or by repairing the effects of
out of order events to give the illusion that they were executed serially. A happened-before
relation can be defined such that events within a single sequential process are ordered, and events
116
sent between processes are ordered by a causality relation. A total ordering of events can be
determined through the use of either a centralized broadcast server, or through the use of a
logical clock and unique global timestamps.
A centralized server, while simpler, can quickly become a bottleneck, and reduce
scalability and reliability. To avoid this, a replicated approach can be used. Global timestamps
ensure that all events have a unique time at which they occurred, allowing for an ordering to be
determined. A logical clock, which is advanced when events are executed, can be used to ensure
that causality relations are preserved.
Non-optimistic, or conservative, serialization ensures that events are always executed in
the correct order at all sites by waiting for previous events to arrive before executing a given
event. This restriction can reduce performance, since not all events need to be executed in
absolute order. Optimistic, or aggressive, serialization is based on the assumption that
conflicting events are rarely received out of order, and that it is more efficient to execute events
as received and occasionally repair problems.
Repair of conflicting events can be accomplished through either rollback or
transformation. Rollback reverts the system to its state just before the out-of-order event
happened. The event may then be executed, and the system brought forward again.
Transformation attempts to transform the out-of-order event using a set of rules such that its
effect is the same as if it had arrived in order.
The use of global timestamps for concurrency control guarantees serialization. However,
the implementation of such an approach is often complex, and can lead to strange behaviour
from the user's point of view depending on the level of optimism. In fact, as Ellis and Gibbs
(1989) argue, it is questionable whether complete serialization is always sensible for interactive
collaborative applications. The acceptability of serialization also depends on the granularity of
the interactions. When working at the pixel level, for instance, users tend not to notice the
unexpected effects of serialization, whereas at a coarser level the undoing and redoing of actions
can be disorienting.
The level of optimism used in serialization also effects the responsiveness of the
interface. Non-optimistic serialization blocks event handling until there is a guarantee that
events are going to be executed in a global order. A user may have to wait for an action to be
accepted, a delay which can break the flow of interaction. With optimistic serialization, there are
no delays; when events arrive late, the system can either undo executed events or transform the
117
new event to resolve the ordering problem. Both these solutions can have strange effects from
the user's perspective.
Researchers have tended to use customized optimistic serialization algorithms. For
example, GROVE (Ellis and Gibbs, 1989) uses a distributed operational transformation
algorithm to transform operations that arrive out of order. This is not a strict serialization; the
system tries to maintain a reasonable order that avoids conflicts between events based on the
semantics of the application. Another approach to serialization is GroupDesign (Karsenty and
Beaudouin-Lafon, 1993) , which manages serialization by allowing commutative events to be
executed out of order, and then using undo/redo to reorder non-commutative events. The
algorithm tries to minimize the need for undos and redoes by detecting masking and
commutativity between events.
Locking
Another approach to concurrency control is the use of a locking mechanism to give a
process mutually exclusive access to a data object. Locks can be shared, allowing several
readers and one writer, or exclusive, allowing only one process to access the object. Locks are
requested when a process wants to access a resource. If no one else holds the lock, the request is
granted, otherwise it is denied. When the process no longer needs the resource, the lock is
released. Locks may be managed in a variety of ways, ranging from a centralized lock manager
that controls resources at one site, to replicated lock managers that require a majority of sites to
agree on a lock request.
As with serialization, locking protocols may be optimistic or non-optimistic. Non-
optimistic locks require that a requesting site wait until the lock is granted before acting on the
resource. Semi-optimistic locks allow the site to get a tentative lock, and take action on the
resource; however, the lock may not be released until a reply is received. If the lock is granted,
the action is committed, otherwise it is aborted. Fully-optimistic locks allow the site to release
the lock and possibly make additional lock requests. If the lock is eventually denied, the site
must perform recovery in the form of rollbacks or undos on the denied resources.
The granularity of locking influences the style of collaboration encouraged by an
application. Very coarse-grained locking leads to a turn-taking protocol, where only one user
can access the shared workspace at a time. This is often used for collaboration-transparent
systems, where an existing single-user application is shared between several users who can take
turns providing input (Lauwers and Lantz, 1990) . Locking can be set at a finer level, increasing
the richness of interaction but also increasing the network traffic. The appropriate granularity
118
depends on the application and the specific task. Aspects provided a paragraph-level locking
mechanism, which afforded the scribe approach to collaborative writing. A much finer
granularity of access control, such as selection or character level locking, would allow users to
work much closer together. However, this may lead to a more anarchic process of writing, with
little coordination of the writers’ contributions. The appropriate level of locking depends on the
group and the type of writing being performed. The interface should allow the appropriate level
to be used, rather than imposing a specific work style.
The way in which users are forced to wait for locks also has an impact on the interface.
With non-optimistic locks, users either have to wait for a lock to be granted, or are denied access.
This is acceptable for atomic transactions in a database system, for example, but delays can be a
problem if they are noticeable to the user. Lags will interfere with the flow of interaction,
making the system seem sluggish. This would be especially annoying when composing text.
The point at which locks are granted can accommodate this lag. For example, locks could be
requested when a user makes a selection, rather than when a character is entered, avoiding delays
during text entry. However, this could prevent other users from accessing that same text if the
user with the lock does not intend to actually alter the selected text. Another alternative would
be to request a lock when typing starts, but this may incur an unacceptable delay in startup.
The alternative is optimistic locking, which avoids delays but presents challenging
implementation and interface design problems. When a lock is denied, the system needs to
revert to a previous state. With fully-optimistic locking, the user may have accumulated a
sequence of tentative locks, so the system needs to decide what to do when a rollback is needed.
This can be annoying to the user, and is possibly difficult if several denied actions are
interdependent. With semi-optimistic locking, the user cannot go on to another action until the
lock transaction is completed. As with non-optimistic locking, the interface could become jerky
and uneven. However, denial is easier for the user to understand, since only one action is
involved. The main difficulty is how to show the user that the action is tentative, and when to
propagate tentative actions to other users.
As an example, the SASSE collaborative editor (Baecker, Nastos, Posner and Mawby,
1993) provides non-optimistic locking at the character level. This can introduce delays waiting
for lock requests whenever users change their insertion point, but allows users to work very
closely, and with immediate response once a lock is granted. Aspects provides non-optimistic
locking at the paragraph level, which reduces delays but does not allow users to work as closely.
From the experience of the Prejudice Project it has been seen that users tend to work at varying
granularities depending on the specific task, such as brainstorming, composing and editing, and
119
depending on the approach taken to writing and the roles of the group members. Joint writing
requires a very fine granularity, whereas scribe and parallel writing both require rapid
notification, but can be accomplished at a coarser granularity of access. This questions the use of
one locking scheme, suggesting that more flexibility is required in terms of granularity.
Concurrency Control in CalliopeThe above discussion reveals the difficulty of choosing an appropriate method for
guaranteeing consistency in the shared text workspace during concurrent access by several
authors. The methods that are technically easiest tend to involve minimization of concurrent
access, which is reasonable for independent writing, and may be appropriate at a certain level for
parallel or scribe writing, but not for joint writing. Conversely, to support the fine-grained
access necessary for joint writing requires the use of highly specialized and complex algorithms.
Issues to address when developing mechanisms for concurrent access to the shared
workspace include notification, responsiveness, and the impact of different access control
mechanisms on the group. Notification should be maximized when the group is working closely,
and can be relaxed otherwise. Responsiveness should generally be maximized, except where
delay in appropriate, such as when explicitly requesting a lock on a region of text. Delay should
be a natural part of interaction, rather than a distraction, and should be made completely visible
to the user. Access control can range from no control, involving the use of serialization to
maintain consistency, to full control with minimized concurrency.
The control mechanism should allow flexibility, with different levels depending on
context. The control mechanism should be dictated by the writing approach, rather than the
writing approach being dictated by the control mechanism. However, it is not likely that the
group will be able to choose the right concurrency mechanism without an explicit understanding
of the mechanism’s impact on the group’s interactions. The system should allow smooth
movement between different writing styles, reducing the costs of speaker change and startup.
The concurrency mechanism should provide the appropriate levels of notification and control for
the task being performed. It is important that the system not impose patterns on the group’s
natural social interactions.
In Calliope, I decided to make the full range of concurrency control mechanisms
available to the system, including both serialization and locking. Locking is available at the
selection, word, line, paragraph and document level. This implementation of mechanism, rather
than policy, at the system level makes the issue of how to choose the correct concurrency control
120
mechanism an interface design rather than a system design issue. Calliope provides a number of
different shared task spaces, which take advantage of the full range of concurrency control
mechanisms provided. I will discuss these spaces in chapter 11. Here I will focus on the
technical issues involved in the implementation of the different mechanisms.
Concurrent access with guaranteed consistency is a strict requirement for a collaborative
editor. It is essential that the text remain the same on all users’ screens. The text need not
always be exactly the same at a given instant, but, depending on the rate at which the system
notifies users, there must be a guarantee of convergence of the contents of the document when
the system enters a rest state (Ellis and Gibbs, 1989) . This consistency can be maintained by
either serializing operations to ensure that changes to the document contents are executed in the
same order, or an equivalent transformation of that order, on all copies of the document. An
equivalent translation is one which results in the same document state.
Calliope implements serialization using a centralized multicast server, and locking using
a combination of a central communications server and replicated locks. The data model is
replicated, with a copy of the text document on each user’s instance of the Calliope application.
Using an open protocol approach (Roseman and Greenberg, 1996) , a mixture of different
locking levels can be used within the same active session. The local system determines the level
of locking to be used locally, and interprets incoming requests. This allows each local copy of
the conference to have a different level of access control.
There are two basic operations that can be performed on text in Calliope: select and
insert. Select changes a user’s selection to cover a specified range of text. Insert replaces the
text in a user’s current selection with a specified text string. Select does not change the contents
of the document, whereas insert does. For locking concurrency control, there is a third
operation, a lock request, which requests write access for a user on a specified range of text. As
with a selection, this operation does not alter the document contents. It is important that these
operations be applied consistently across all copies of the document.
For serialized concurrency control, operations are simply serialized through the central
server, and then executed on each copy of the document. This ensures consistency, but can lead
to delays, especially if the network connection between the server application and the other
applications is slow. The alternative is to use locking, which allows local insert operations to be
performed before being sent to the rest of the copies of the application.
Consistency is maintained using a mechanism similar to that used in ShrEdit (Olson,
Olson, Storrøsten and Carter, 1992) and SASSE (Baecker, Nastos, Posner and Mawby, 1993) .
121
When a lock request is made, the request is serialized through the central server. This ensures
that all lock requests are seen in the same order on all copies of the conference. When a request
is received, the local user checks to see if the requested region is free. If it is, the lock is granted.
Similarly, selection requests which do not involve a lock are serialized through the server.
When an insert is performed, it is executed first locally, and then sent to all other copies
of the conference through the serialization server. Performing the insert locally and then
transmitting it to the network maximizes local responsiveness. When the operation is performed,
an operation vector is incremented for each user in the session. This vector keeps track of the
number of changes performed by each user on the local copy of the document. Since insert
operations are only permitted on regions on which the local user has a lock, there is no
possibility of conflict. The operation vector is used to translate incoming selection and lock
requests. Since local insert operations could have been performed that the remote user does not
know about, the selection must be transformed to take into consideration these local inserts.
Similarly, lock and selection ranges stored in the local document must be translated when
changes are made to the document.
Currently, Calliope provides a main text window with immediate notification and user-
selectable levels of locking. Any combination of locking levels are available, and can be chosen
from a menu item (see Figure 9.1). The system also provides private text, and a number of other
shared spaces with varying levels of access and notification, which will be discussed in chapter
11.
Figure 9.1: User-selectable locking levels.
Within the main text, the impact of the control mechanism can be reduced by ensuring
visibility of the access control system. When users have a lock on a text region, this must be
122
clearly shown; it should also be obvious to the user whether she has write access or not. In
Calliope, this is accomplished through the use of two interface features (see Figure 9.2). When a
user has a lock on a region of text, the text is outlined with a colour-coded box. This box shows
at a glance that the text is locked, and by whom. In addition, the local user’s cursor displays an
indicator informing the local user whether or not she has write access to the selected text.
Figure 9.2: The main text window in Calliope.The figure shows selection-level locking. The local user, displaying a padlocked cursor, is locked out.
By providing a variety of concurrency control mechanisms, Calliope allows the group,
rather than the system, to determine the way in which it will write together, using any one of the
approaches to synchronous collaborative writing. When the group wants to work closely, using
the joint writing approach, they can choose to use selection or word level locking. If the group
members want to work using the scribe approach, with only one person accessing a section of
text, or if they want to work using parallel or independent writing, they can switch to paragraph
locking. The problem with this design is that it relies on the group understanding the affordances
of the different levels of locking. The alternative would be to have the system adapt to the way
in which the group is working. Future work should address this issue.
This chapter has focused on the basic mechanisms necessary to provide resource sharing
in a shared text editor. In addition to the sharing of the text document, a collaborative writing
123
tool also needs to provide support for awareness of the presence and actions of collaborators in
the shared workspace, and the history and context of the collaboration which led to the
development of the text document. In the next chapter I will discuss the requirements for
supporting awareness in collaborative writing, and describe the mechanisms provided by
Calliope to satisfy these requirements.
125
Chapter 10
AwarenessAwareness of the presence and actions of collaborators can be encouraged in a number of
different ways. These approaches can be divided into tools which provide peripheral awareness,
and those which support a sense of history and context of the ongoing collaboration and the
developing shared text. Peripheral awareness involves keeping all users aware of the fact that
they are working in a shared workspace, and providing ongoing, non-intrusive notification of the
actions and locations of all other users. The system should also keep information available about
the rest of the group and the state of the shared text, providing a shared context to help encourage
communication and cooperation.
Chalfonte, Fish and Kraut (1991) argue that the ease and success of collaboration
depends on the communication tools provided by a system. Interactivity and expressiveness are
two dimensions which can be used to distinguish rich channels from impoverished channels.
Chalfonte suggests that text affords a focus on the content of communication whereas speaking
affords a focus on the conversational partner. Computer support for collaboration should allow
for the focus to shift smoothly between task and communication.
Buxton (1992) divides the space between collaborators into a “person space” and a “task
space”. The person space encompasses the collective sense of copresence between and among
the group of participants in a collaborative activity. Task space is the notion of copresence in the
domain of the task being undertaken. Buxton sees the goal of technological support for
collaboration to be “seamlessness”, or the transparent bridging, of these two spaces. This can be
accomplished by allowing collaborators to move between person and task space without any
change in the skills necessary to communicate. The task space should support awareness of the
presence and actions of collaborators, essentially extending the person space. For example, a
collaborative system should provide a shared person space even when the group’s focus of visual
attention is on the computer screen, as well as when the focus is on the collaborators. Buxton
claims that “some (many or most?) complex tasks require a range of channels and modalities of
communication in order to be effectively supported.” (p. 820) From the Prejudice Project
analysis I have shown how the presence of the shared workspace impacts the distribution of
attention between group members, and how communication can come to be grounded in the task
space. The system should reduce the cost of switching contexts between person and task space,
and encourage users to engage each other, not the system. While most systems focus on either
person space or task space, some recent work, such as the development of ClearBoard (Ishii,
Minoru and Grudin, 1992) , has begun to address the need to bridge the two spaces.
126
This discussion stresses the importance of maintaining an awareness of both the shared
space in which you are working, and the fact that there are other people working in that shared
space. Dourish and Bellotti (1992) indicate that the use of shared feedback, the notion of
providing implicit, peripheral information about everyone in a shared space, is a promising
approach to providing mutual awareness. Examples of this approach can be seen in ShrEdit
(McLaughlin Hymes and Olson, 1992) and SASSE (Mawby, 1991; Nastos, 1992) . ShrEdit
allows you to request that the system find and track movements of others, but provides no
continuous feedback to users to indicate the locations of other users’ editing activities. SASSE
provides peripheral information in the form of colour-coded, shared scrollbars, audio cues, a
document overview (or gestalt view), and a tracking mode. At the same time, users need to be
able to explicitly request information about the context and content of the shared workspace,
including information about who made contributions and changes to the document and when.
Calliope provides a number of awareness mechanisms. In the main text window, colour-
coded selections, telepointers, and shared scrollbars provide awareness of the location and
actions of others directly in the focus of the user’s attention. The shared scrollbars show user
locations, and allow the local user to synchronize her view to that of a given remote user, for
optional WYSIWIS view sharing. An overview window provides additional peripheral
information, combining fisheye point-of-view displays for each remote user and a gestalt-like
overview. History and context information is stored for all the text in the main window; text can
be displayed colour-coded to indicate who entered the text. Text can also be queried for author,
access time, and other historical information.
Peripheral AwarenessParticipant activity is a valuable source of information. In face-to-face conversation
information about participant activity is maximized. Participants will be aware of the
interactions that led to the present utterance, and will know the content of the utterance and from
whom it originated. A collaborative editor should attempt to give collaborators using the
computer system a similar degree of awareness of participant activity. The tendency of the
shared workspace to become a focus of the group’s attention makes this need for awareness
within the task space especially important, even in a co-located situation.
Awareness can help to reduce conflict and the reliance of the system on concurrency
control mechanisms. According to Hughes (1993) , a shared editor needs to enable users to
avoid unintentional conflict, handle situated conflict, and resolve conflict. This should be done
by means of an appropriate mechanism to provide the necessary information and awareness.
What is needed is “social WYSIWIS” or a system that inspires trust that all participants can,
127
indeed, see the same shared object, and leads participants to treat the shared workspace as a
common object. Mutual awareness and enhanced copresence and visibility obviates the need for
extensive control mechanisms. The concern should be for user communication and the
coordination of views and shared knowledge more than for explicit coordination in the form of
locking or floor control.
A system should provide peripheral awareness of the presence of both the local user and
others in the shared workspace. The goal should be to make the shared nature of the workspace
obvious to users, and encourage communication between collaborators. Peripheral awareness
includes awareness of both the actions and potential actions in the workspace. The system
should provide information about who made changes to the text, when, and why. It should also
provide information about what others are doing, and how those actions affect the local user.
Conversely, the system should provide information about the local user, showing how her
actions can affect and have affected others. It should be obvious who is doing what and who did
what, to make conflict and potential conflict visible to the entire group.
Shared Feedback
Calliope provides shared feedback by allowing users to display remote cursors on all
collaborators’ screens, by showing positional information via shared scrollbars, and by allowing
users to follow others’ actions by means of an optional What You See Is What I See (WYSIWIS)
mode.
The ability to have a user’s cursor position show up on all users screens, by means of a
telepointer mechanism (see Figure 10.1), provides two types of awareness. It peripherally makes
users’ actions and movements visible to all, encouraging the sense of a shared context. It also
allows users to refer to the shared workspace by means of gestures. Telepointers are small, so as
to be non intrusive, and are colour-coded.
Figure 10.1: Telepointer and shared scrollbars.
128
Further awareness of the location of remote users is provided by a set of shared scrollbars
(see Figure 10.1). Shared scrollbars are displayed parallel to the regular local scrollbars, and
show the relative position of all users in the main text workspace. As with the telepointers, the
shared scrollbars are colour-coded. Each user has a unique colour, which is used consistently
throughout Calliope to identify a specific user. This lets users know at a glance who is
performing an action, and where the members of the group are looking in the document.
In addition to knowing where a user is in the document, it is often useful to be able to see
exactly what a given user is doing, especially when the group is using the scribe writing
approach, or during editing of the document. To support this need for close awareness of a
specific user’s actions, Calliope allows a user to optionally link her view of the main text to that
of any other user (see Figure 10.2). Once the views are linked, any scrolling performed by the
chosen remote user is mirrored in the local user’s document. The local user is also free to unlink
the views at any time and to scroll elsewhere. The remote user’s view is unaffected by the link.
Figure 10.2: Following a remote user’s view.The local user has clicked on a remote scrollbar, and has the option of linking her view with the remote user.
Overview Display
The inclusion of telepointers, shared scrollbars, and an optional WYSIWIS view provide
a sense of the presence of others, and a certain degree of peripheral awareness. Additional
awareness is provided by an overview window, which combines two important ideas. When
working on a shared document, it can be very disorienting trying to visualize the entire
document, and the location and actions of all collaborators within that document. To provide a
sense of the structure and content of the entire document, an overview or gestalt view can be
used (Nastos, 1992) . This view can be used to see, at a glance, how the text is changing, and
where people are in relation to each other and the overall document.
129
Another important piece of information that should be made available to the user is what
others are doing, who they’re working with, and how they are changing the document. This can
be addressed by providing a point-of-view window on a user’s work (Gutwin, Stark and
Greenberg, 1995) . However, this sort of view may not always be needed, and uses up valuable
screen real-estate. Calliope deals with this problem by combining the idea of a gestalt view and
a point-of-view display into an overview window with fisheye lens displays of each user’s
context and actions (see Figure 10.3). This provides an overall view of the content and structure
of the document, as well as the context and actions of each user, at a glance.
Figure 10.3: Fisheye overview display.The local user is working at the top of the document, and is watching a remote user work at the end of the document.
The main problem with this approach is that the fisheye lens interferes with the overview
window’s display of the structure of the main document. This can be minimized through careful
selection of the amount of distortion applied to the fisheye lens. One way to deal with this is to
allow the local user to select the amount of zoom. If desired, the user could flatten the fisheye
lens, so that the overview displays a true image of the document structure. Location and actions
of users would still be displayed by colour selections and changes to the text. If the user is
interested in a specific user’s actions, the fisheye lens could be magnified at that location. The
choice of the default magnification could be a compromise between flattened text and full
130
magnification, to maximize the display of overall structure while affording some notion of the
context of each user.
History and ContextAnother problem introduced by collaborative tools is the need to keep track of changes in
a shared document. Information should be stored with the shared workspace which shows the
development of the shared text and the connection between this history and the communication
which has taken place around the text. The embedding of history and context in the shared
workspace affords reviewability. When the context of a text is unclear, there is potential for that
text to be misunderstood. If users are able to delete text irrevocably, then the context of the
interaction, which structures the group process, is lost. One way to provide the necessary
information about changes in the document is through the display of differences in the document,
often called "diffs", either as change bars (Neuwirth, Chandhok, Kaufer, Erion, Morris and
Miller, 1992) , through the use of annotations (Nastos, 1992) , or through more active
notification (Minör and Magnuson, 1993) .
The user should be able to ask questions such as “who did this?”, “why was this text
changed?” and “when was this paragraph last edited?”. This requires that the system keep some
form of record of the change history of a text, including ownership and access information.
Calliope provides this information in the form of coloured text; as users enter text, the
information about who entered the text is stored as colour formatting (see Figure 10.4). This
formatting can be toggled on and off by each user on their local view of the document. The
coloured format information shows who made changes at a glance, providing clues as to who has
contributed to given portions of the document.
Figure 10.4: Coloured text showing who wrote what.
131
An important consideration is whether this colour tagging should be permanent. As a
text is developed, the group’s notion of ownership can change drastically. It isn’t clear at what
point an item of text becomes “public”, nor is it clear who should be given credit for the creation
of an item of text. The act of entering the text into the workspace does not necessarily imply
emotional or intellectual ownership or credit for the content of the text. The group must be able
to adapt the information stored in the document to fit their model of the text.
From my observations, I have seen that groups tend to change their notion of ownership
of text as a document develops. When first entering text, there is a connection between who
typed the text and who is identified as owning that text. However, as the document is altered and
edited, the group, ideally, begins to think of the text as belonging to the group, as opposed to any
one individual. At some point, the colouring on text should be removed, to reflect the fact that it
is no longer important who wrote a specific text fragment. This could be done by means of a
voting mechanism, or some sort of consensus or approval mechanism that requires that the entire
group agree on the movement of the text from ownership by a single user to ownership by the
entire group. Calliope approaches this by leaving the decision-making to social protocols. Any
group member can remove text colouring at any time; these changes are immediately visible to
all users, encouraging discussion and negotiation if any group members disagree with the
decision.
Figure 10.5: Embedded author information.
The colouring of text based on author provides awareness of the history of a text
document. The system should also let users ask questions about the history and context of
specific text fragments, or about the entire document. Calliope supplements the colour-coding of
text by allowing users to query a text fragment for information about who wrote the text (see
Figure 10.5). This could be enhanced by storing information about the previous text at that
location, when the text was changed, and contextual information such as why the text was
changed.
132
In this chapter, I have discussed the ways in which a synchronous collaborative editor can
provide awareness of the presence and actions of others in the shared workspace, and how the
system can support a sense of history and context of the ongoing discourse within the shared text
document. In the next chapter, I discuss the requirements for the design of the shared task spaces
in which the group works to write together, and examine how Calliope explores these design
requirements.
133
Chapter 11
Shared Task SpacesThere are several issues to consider when designing the shared spaces provided by a
collaborative writing system. The most important consideration is the character of the main text
window, which serves as a focus of the group’s discourse. Careful consideration must be given
to how to deal with the questions of sharing, access control, and notification of changes. There is
also an issue of what model to use for the shared document. A straightforward representation of
the text as a linear text document with paragraphs is the most intuitive, but tends to impose a
certain style on writing; alternative approaches include an outline format, an unstructured
scratchpad, a brainstorming tool, or a hypertext.
Synchronous collaborative editors have tended to focus on the problem of providing a
shared context, giving users a common shared document in which to work and around which to
discuss their composition. While some systems, such as GROVE (Ellis and Gibbs, 1989) and
SASSE (Baecker, Nastos, Posner and Mawby, 1993) , have provided alternative views of the
shared text, all systems have remained within the model of a single text space. As has been
shown in the discussion of the Prejudice Project, the provision of a single text space, when
combined with a coarse-grained access control mechanism, can lead to dominance of the text and
the discussion by a single member of the group.
The requirements for resource sharing suggest that no single concurrency control
mechanism is appropriate for all stages of the writing process, or even for all group members at a
given time. This suggests that a synchronous collaborative writing system should provide a
range of shared task spaces in which the group can work, with views on the main text and
alternative text spaces that can be chosen as appropriate. The central document, which is the
focus of the group’s task of writing together, gives the group a common focus and context, and is
the point around which the group work revolves. Additional views and shared spaces can help to
reduce the costs of grounding, provide multiple related channels of communication, and afford
smooth movement between the various approaches to writing together.
Calliope supports a number of different shared task spaces in which the group can work
and interact. The main text window provides the focus of the interaction, and is where the
shared document is developed. In addition, Calliope includes private text windows, an
unstructured scratchpad, metatextual annotations, and external annotations. I will now discuss in
detail each of these shared spaces, and explain how they satisfy the design requirements derived
from the analysis of the Prejudice Project.
134
Reduce Costs of GroundingThe access control mechanism provided by Aspects afforded the scribe writing approach.
Mantei (1988) suggests that in a system where participants have to rely on a scribe to enter
ideas, the participants tend to get frustrated when they need to communicate large amounts of
information to the scribe. The ability to express ideas directly might reduce conflict. At the
same time, if several writers are allowed to access the same text concurrently, there is no order to
the construction of the text, and the result can be chaotic. It is sometimes useful to have one
person control the text and direct its composition. However, it is also beneficial to allow all
group members equal ability to contribute ideas to the developing text. This suggests that
various different levels of access control and document structure are appropriate at different
points in the process of collaboration. The system should provide a variety of access levels to
support the different writing styles. This would reduce the costs of speaker change, start-up, and
formulation, production and repair of textual utterances.
The system should also provide a variety of levels of sharing. As users compose text, it
is sometimes useful for the text to be immediately visible, to encourage comments and increase
the sense of a shared context. This can be thought of as a way of communicating through the
artifact. Participants in a synchronous interaction will be able to watch others’ text as it is
created. However, one of the benefits of written communication is the ability to polish ideas
before sharing them. Fine grain text transmission mitigates against one of the advantages of
written communication, the ability for the author to review and refine her thoughts before
transmission. This suggests that the notification of text entry should be flexible enough to allow
users to choose how public or private the text will be, and allow the level of privacy to be
changed as desired, to reduce the cost incurred by lack of delay. By careful choice of the way in
which the text spaces are shared and by providing flexibility in terms of access to the shared text,
the system can be designed to reduce the costs of grounding in communication about the
developing shared text.
Different “Windows” into the Main Text
The design of the main text space must take into consideration a number of requirements.
The need to facilitate moving between different writing styles, such as joint, scribe and parallel,
means that there must be flexibility in terms of access control and notification. The group has to
be able to work more or less closely, without interference from the system’s concurrency control
mechanism.
One way to deal with the need to have different levels of access control and notification
is to provide a number of “windows” on the main text, each with a potentially different level of
135
access control. When you want to work on a given section of text, you can choose how often
your changes are sent to others and how closely to you others can work. This window could be
an explicitly separate window on the user’s screen, or could simply be a region within the main
text window. Within this text window, the level of access control and notification would be set
as appropriate to the type of work being performed. One difficulty with this approach is how to
choose the level of access control and notification. This could be done automatically by sensing
how closely people are working, or selected explicitly by the user. Automatic selection is
difficult, and has been seen to interfere with the social interaction within a group. However, user
selection imposes an additional cognitive load on the users. More work needs to be done to
address this problem.
Figure 11.1: Combining different locking levels.The local user has word level locking; the remote user has paragraph level locking.
Access can range from free-for-all, with serialization, to floor-control where only one
user can access the text at a given time. Notification can range from immediate, for close work,
to completely private text that is made public only when the local user chooses. Currently,
Calliope provides a main text window with immediate notification, and user-selectable locking
levels (see Figure 11.1). The system also provides private text, and a number of other shared
spaces with varying levels of access and notification, which will be discussed below.
Private Text Windows
One of the advantages of text-based communication is the ability to refine and polish
ideas before transmission. This suggests that the shared task spaces provided by a collaborative
editor should allow for private entry and editing of text. However, it is important to retain the
shared context of the common task space, and allow free movement between the public and
private spaces. Calliope provides the ability to enter text in a private window, to let ideas to be
formed and polished before being exposed to group and thus minimize the cost of the lack of
delay in transmission.
136
There are two important issues to address when providing a private text space: how to
afford seamless movement between the private and public spaces, and how to handle awareness
and notification.
The first issue can be examined from two perspectives. First, the user must be able to
easily move from the main text space to a private text space when she wants to work on a given
fragment of text without immediately letting the changes be seen by the rest of the group. The
move from public to private should not involve too much overhead, or the user will not bother to
move to the private space. The effort must also not interfere with the flow of creativity during
writing. In Calliope, this is handled by allowing the user to move the text currently selected in
the main window into a private text window with a single mouse click. The text can then be
edited freely in private (see Figure 11.2).
Figure 11.2: Private text.
The other side of the problem of movement between public and private is the issue of
moving back into the public space. Calliope allows the user to move the edited text back into the
selection in the main text space with a single button press. In combination with the selection-
level locking mechanism, this allows users to choose to work on a given section of text in
private, and notify the rest of the group when desired. An alternative to simply moving the text
back into the main text would be to allow the user to choose to update the main text but continue
137
working in private. The nature of the private space must also be considered. In Calliope, the
private text space takes the form of a separate window; it might be easier on the user if the text is
edited in-line, but in private. However, the use of a separate private text window helps to
indicate to the user that the space is distinct from the main text.
The second issue in the design of a private text space is the question of awareness and
notification. Notification was mentioned above in the discussion of moving between spaces.
However, in terms of awareness, the rest of the group should be kept aware of the fact that a user
is working on the text in a private window. This will encourage communication about the text,
and reduce the possibility of conflicting changes to the text. Ellis, Gibbs and Rein (1991)
suggest that, when users are working relatively independently, they do not need immediate
notification of others’ changes. They suggest the use of “cloudburst” indicators to show the
region where a remote user is making a change. When the local user stops making changes, the
system updates remote changes. This approach could be applied to Calliope. When a user has
moved a region of text into a private space, cues such as colour change or stippling could be used
to indicate potential change. If the text is actually entered, then the change is propagated to all
users. If the change is cancelled, then the text is reverted to its normal appearance.
The Scratchpad: An Unstructured Text Workspace
One of the sources of the problem of access and notification is the fact that, in traditional
collaborative editors, all interaction and writing takes place in the main, “official” text
workspace. This imposes certain assumptions and structure on textual utterances. The design of
a shared editor can also include experimentation with different spaces, each providing different
affordances in terms of task and communication. For example, it is possible to provide shared
text workspaces with different levels of structure, altering the methods used for access and
modification, and changing the assumptions about the role of the text in the group’s performance
of the task of writing together. To afford contributions by all, the system could provide a more
unstructured workspace which allows entry and modification of ideas by all participants.
To explore this concept, Calliope provides a scratchpad, an unformatted whiteboard-like
shared window that allows text to be entered and manipulated freely (see Figure 11.3). The
scratchpad also allows the text to be gestured at with telepointers, and marked up with freehand
annotations. Text can be dragged or cut and paste between the scratchpad and the main text
window. The scratchpad is intended to reduce the cost of entering ideas for non-dominant group
members, and provide multiple focuses for discourse, rather than just the end of the current
paragraph in the main text workspace. Group members can put down ideas whenever they want,
and all ideas will be of equal status, since there is no predetermined structure to the document.
138
Figure 11.3: The scratchpad: an unstructured text space.
Other systems that attempt to support the use of an unstructured space, such as Cognoter
(Tatar, Foster and Bobrow, 1991) , have run into difficulties, due to the use of a parcel-post view
of the communication that takes place using an electronic shared workspace. The results from
the use of Cognoter suggest that a more interactive model of communication, such as that
observed during the Prejudice Project, is more appropriate. To support this model of
communication, the scratchpad is designed to provide immediate notification of changes, and to
support referentiality and the sense of a shared context to aid users in coordinating their
conversational actions. Text items are immediately visible, and can be edited by anyone. To
prevent consistency problems, locking is provided at the text item level, with locks clearly
indicated by coloured selection indicators.
Participants will use various means to refer to particular artifacts, attempting to establish
mutual references to artifacts (Miles, McCarthy, Dix, Harrison and Monk, 1993) . Gestures can
be one means used to ground references, and help to reduce the cost of display. Gestures can be
supported through the use of telepointers; further, the ability to informally annotate text through
the use of gestural annotations or markups can help to link direct and indirect communication.
These gestures should be supported in any of the various shared text spaces provided by the
system. The scratchpad provides support for gestures and freehand annotation; the other spaces
in Calliope currently only support gesturing with telepointers.
139
The fact that the scratchpad is separate from the main text workspace encourages
experimentation with ideas and content. Since it is a separate, less “official” workspace than the
main text, it is easier to enter alternative ideas and wordings, and play around with different
approaches to writing the text. However, the fact that the scratchpad is a separate workspace
also introduces the problem of grounding the ideas and text entered in the scratchpad in the main
text. There should be a connection between these experimental ideas and those entered in the
main text window, and there should be a relatively seamless way to move ideas and text back
and forth between the scratchpad and the main text window. This could be provided by allowing
annotative links between the scratchpad and the main text, or by allowing easy cut and paste or
drag and drop between the spaces. Calliope begins to deal with this problem by allowing cut and
paste between the scratchpad and the main text.
An unstructured workspace is not the only way in which to encourage the free entering
and discussion of ideas and text fragments. An outline view or a brainstorming tool could also
be used to allow all users to enter ideas. However, an outline view, while having less of a single
focal point than a normal text document, still imposes a rigid structure on the workspace. Users
can enter ideas at any point in the hierarchy, but only within the hierarchy. A brainstorming tool
allows ideas to be entered in any order, at any time, but does not allow for manipulation of or
direct reference to the shared workspace. The scratchpad adds the ability to manipulate ideas
freely, allowing the workspace to become a more integrated part of the discourse around the
developing text.
Multiple Related Channels of CommunicationThe communication that takes place during collaborative writing can be divided into
direct and indirect communication (Miles, McCarthy, Dix, Harrison and Monk,, 1993) . Direct
communication implies that participants are aware of each other, and will address messages to
their co-participant(s). Successful direct communication implies the achievement of a common
understanding between participants. If each participant can successfully communicate her
perspective, a common understanding can be achieved, with each participant having some
knowledge of how other’s perspectives differ from their own. Understanding need not only be
an emergent property of direct communication, but may also come from interaction through
shared artifacts. Cooperative work emphasizes some shared task, or common purpose. Often
this shared task will involve some artifacts that are the subject of the work. In addition to each
participant’s interaction with the artifact, communication can take place through the artifact, in
the form of indirect communication.
140
The main channel of indirect communication in a collaborative editor is the shared text
workspace. This provides a shared context for the group, and a space in which ideas can be
constructed and negotiated. However, the use of a single, structured text as the only channel of
indirect communication limits the group to a certain form of communication. Participants’
perception of a communication channel will be influenced by its appearance, and the way they
interact with it. According to Lea (1991) , users “actively construct for themselves a view of the
medium that is uniquely contextualized by their interpretations of the social and communication
worlds that they occupy.” (p. 160). When a channel is perceived as supporting a particular type
of communication, then that channel will help to specify the intended meaning of messages
received on it. McCarthy, Miles and Monk (1991) look at the difficulties of achieving common
ground in text-based communication, suggesting that a common report space separate from the
discussion itself would add to the ease of communication. A space separate from the main text
but related in some way, by pointers or hypertext links, could provide this sort of external
common space.
Annotative communication is a common means for exchanging ideas in collaborative
writing. This form of communication can be very powerful, allowing users to freely contribute
to the discussion and the formation of the text, without actually having to be able to modify the
main text. Receiving related information on more than one channel can have a strengthening
effect. This also allows for deictic referencing (i.e. “this text” or “that paragraph”), which can
increase the efficiency of communication. This implies the need for some means of linking the
various spaces. The fact that annotations directly refer to the text grounds the contents of the
annotation in the ongoing discourse.
Metatextual references or annotations are common in asynchronous collaborative writing
systems. In PREP (Neuwirth, Kaufer, Chandhok and Morris, 1990) , annotations and draft-
passing are the only form of collaborative writing. Quilt (Leland, Fish and Kraut, 1988) takes
communication into consideration in the form of annotations and an integrated message passing
system. The use of annotations provides a level of metatext which allows communication about
the text without actually altering the text. This support for communication in asynchronous
systems can help to suggest how communication could be more adequately supported in
synchronous collaborative writing tools.
“Live” Annotations
There are a number of issues involved in the design of a shared annotative text space.
Should annotations be visible to all group members? Should they be editable by group members
other than the creator? This could lead to discussions taking place in the note, much like
141
discussion notes in CSILE (Scardamalia and Bereiter, 1992) . Can annotations be deleted? Also,
what happens to orphaned notes, annotations for which the main text to which they refer has
been deleted? Should annotations always be visible in the text and on the screen? How should
they be displayed, directly on the text or in a separate space? All of these issues are important,
and need to be addressed in the design of the system.
Calliope supports the use of annotative communication in the form of text windows that
can be attached to arbitrary sections of text, or to spaces between text, and seen by all members
of the group (see Figure 11.4). These annotations can be edited by all, but only by one person at
a time. This sort of floor-control access affords the message-passing style of interaction that
takes place when collaborators attach notes to copies of a written document. However, the fact
that the notes can immediately be seen by all group members makes the notes a much more
direct form of communication. To avoid the potential cost of immediately visible text, the
annotations can also be created first and then attached to text, using the same mechanism as the
private text space.
Figure 11.4: Annotation window.The figure shows a public text annotation attached to a region of text.
The inclusion of annotative text provides multiple channels of communication, allowing
ideas uttered aloud to be reinforced in text. This use of several forms of communication to
142
contribute an idea allows the idea to have more impact on the group. The use of a textual
medium also gives a more concrete history of the ideas on which the text is built, and encourages
discourse about the text.
External Annotations
The notion of linking comments or annotations to the main text can be extended to
support the idea of reference to external sources or representations of the text. Authors may
want to make a connection between the developing text and source material, to other texts that
they have written at different times or in different groups, or to alternative versions of the main
text (Sharples and Pemberton, 1992) .
Figure 11.5: Attaching a web-based annotation to a point in the text
Calliope provides support for external representations in the form of annotations that,
instead of linking a local text note to the main text, allow users to link pages on the World Wide
Web to the main text by embedding a Universal Resource Locator in the text (see Figure 11.5).
In this chapter, I have discussed a variety of shared task spaces that can be provided by a
collaborative editor. These spaces must be designed to help afford communication by reducing
the costs of grounding in a shared electronic workspace, providing multiple related channels of
communication, and affording the use of the shared text workspace to develop shared
understanding between the group members as they write together.
143
Future Work and Conclusions
In this thesis I have examined the role of communication and the development of shared
understanding in the process of writing together, and explored how computer support for
collaborative writing affects and is affected by the communicative nature of collaborative
writing. Drawing on literature from a diversity of backgrounds, including writing research,
communications research, sociology, computer-supported cooperative work, and media studies, I
looked at how writing together can be considered a form of communication, and shown how this
suggests a new perspective for the investigation of the evaluation and design of shared text
editors. A field study was carried out to study the extended use of a collaborative editor in a
situated context, and the analysis of this study was used to derive a series of design requirements
for a next-generation prototype shared editor. These design requirements were explored through
the implementation of the Calliope prototype shared editor.
Calliope developed out of a number of previous research projects; these began with an in-
depth interview study of how people really write together, which led to the development of
SASE and SASSE, two prototype shared editors. These systems were evaluated in the usability
laboratory, which identified problems and design issues. In my research, the focus of
exploration moved out of the laboratory into the field, looking at extended, situated usage of a
collaborative editor. My analysis of this field study has led to the design and implementation of
a new prototype collaborative writing system.
The field study which provided the basis for my discussion of collaborative writing
involved two groups of grade six students working together to write a magazine on prejudice.
Over the course of the Prejudice Project, the Tuesday and Wednesday groups developed from a
collection of eight students who were arbitrarily grouped together to work in an after-4 program
to a group who identified themselves as the authors of a magazine about prejudice. They saw
themselves as separate from their classmates, having taken part in a research project and written
a magazine which they were proud to identify as their work, and which they felt could make a
difference in people's attitudes towards prejudice and racism (Posner, Mitchell and Baecker, in
press).
Despite the pre-existing tensions and conflicts in the groups, and the different styles and
personalities of the group members, the students did come to identify themselves as a distinct
group. They did learn about prejudice, and were able to create a document which reflected their
shared understanding of the topic. The magazine, a jointly authored document, reflects this
144
shared understanding, and helped lead to the formation of a coherent group. This formation of
the group, and the creation of a common understanding amongst the group members, was
influenced by Aspects, the shared editor used by the group to write their magazine.
The introduction of collaborative technology into the group changed the nature of
interpersonal communication, of control of discourse, and of the negotiation and sharing of ideas
and understanding. This is due to the nature of the medium, the skills needed, and the interaction
of the group members with and around the technology. The introduction of the shared text
workspace also altered the group’s focus of attention. The computer came to act as an
alternative, or sometimes even a primary, focus of the group’s attention. This focus can be used
to avoid giving attention, and as a device for gaining attention. The computer also served as an
additional communication medium, changing the costs of and affordances for grounding of
communication during the development of shared understanding. The use of the computers gave
the group a focus for their collaboration, and provided a medium for their development of shared
understanding.
The observations of the students in the Prejudice Project suggested a number of design
requirements for the development of technology to support collaborative writing. On a systems
level, the support for a shared workspace must provide flexible mechanisms for concurrency
control, allowing concurrent access to the text while maintaining data consistency. The
important position that the shared text document occupies in the group’s focus of attention
requires that a shared editor provide support for awareness of the shared nature of the
workspace, to help to bridge the gap between the shared task and person spaces in which the
group is working. Support for awareness can take the form of peripheral awareness, cues as to
the presence of the group and the actions taking place in the shared workspace. Awareness can
also consist of maintenance of the history and context of the group’s collaboration. The shared
task spaces provided by the system should be designed to provide multiple related channels of
communication, and to reduce the costs of grounding the communication that takes place both
around and through the developing text.
Calliope addresses the design requirements derived from my observations of student
writers during the three months of the Prejudice Project. The system provides a main shared text
workspace, with a variety of access control mechanisms, ranging from selection through to
paragraph level locking. It provides a number of awareness tools, including shared scrollbars,
telepointers, an optional WYSIWIS mode, a fisheye overview display, and text colour-coded by
author. It also provides a number of different windows into the main text and a variety of
parallel shared task spaces, including private text windows, an unstructured text scratchpad,
145
metatextual annotations, and external annotations.
Other systems have attempted to support synchronous collaborative writing. Table 12.1
summarizes what Calliope has added to the development of support for collaborative writing, in
terms of support for task and communication during group writing, as compared to a number of
other synchronous shared editors.
GROVE ShrEdit Aspects SASSE Calliope
Concurrency Control flexible concurrency control + + + + ++
Shared Task Spaces reduce costs of grounding + + + + ++ multiple related channels + - + + ++
Awareness peripheral awareness + + + ++ ++ history and context + - - + ++Notation: ++ system provides good support + system can handle – system does not support
Table 12.1: Comparison of synchronous shared editors based on design requirements.
Calliope begins to address the requirements I have set out as a result of the Prejudice
Project. However, it has been neither formally nor informally evaluated. In addition, the system
is just a prototype and, while exploring some design ideas, only touches on the design space.
The next step to take in the investigation of how to support synchronous collaborative writing is
to begin another iteration of evaluation and redesign.
To evaluate this new system, there are two possible approaches. Evaluation could be
carried out in the laboratory, focusing on specific design problems and interface issues such as
the access control mechanisms or collaborator awareness tools. On the other hand, further
extended usage could be observed in a real task environment. A combination of these
approaches would provide a complementary set of observations which could be used to further
drive the design of collaborative writing systems which support the complexities of group
interaction and social processes in real contexts. An issue that needs to be addressed in terms of
evaluation is how to gather information about the use of a collaborative writing system during
extended, situated usage. In my study, since the students only used Aspects during the weekly
sessions we were running, I was able to capture rich information about system usage and social
interaction. Thought needs to be given to how this type of rich, contextual data could be
gathered when the system is being used by groups writing together on their own time, in various
146
different locations. System event logging can be used to capture simple usage data, but this fails
to reflect the social interactions which take place around the system. One possible solution is to
use the system itself to capture significant events using the multimedia recording capabilities of
modern workstations. Future work should look at how to do this effectively and efficiently.
The implementation of the Calliope prototype collaborative writing system touched on
the design requirements which emerged from my study. However, there are many areas that
require further exploration. Calliope implements a variety of concurrency control mechanisms,
but does not provide a solution to the problem of selecting the appropriate level of control. It
would be interesting to investigate different ways of choosing the type of concurrency control the
system should provide. Should the system attempt to monitor the proximity at which users are
working, and adjust the concurrency control mechanism appropriately? Or should the system
provide various shared task spaces, each with a different level of access control and notification?
There are also a number of interface design issues, including how a system conveys information
about locking and potential conflict, and how users request access to the text.
The various awareness mechanisms provided by the system could also be explored
further. Shared feedback could be extended to help the user see her impact on others through the
use of tools which show how closely others are working to her location and how her changes to
the text affect the group’s environment. These same mechanisms could be used to help the
system determine the appropriate level of concurrency control. Tools could be added to provide
inverse feedback, letting the local user know when others are monitoring her. The design of the
overview could be extended to convey awareness of activity in all of the shared task spaces
provided by the system. The means of displaying overview information could be explored, using
concepts such as a heads-up display or several point-of-view displays rather than a single
overview window.
The display of history and context should also be further investigated. It isn’t clear either
how long ownership information should be stored, or how a decision should be made about
changing or removing this information. There is also the issue of versioning and the
maintenance of the context of changes made to the text. The system could provide information
about the rationale behind changes to the text, in the form of annotations or revision notes, and
could support alternative versions to encourage experimentation.
Work could also be done to explore how the presence of various different shared
workspaces affects grounding and shared understanding. The use of multiple channels of
communication to ground discourse introduces the possibility of discontinuities between the
147
group’s understanding of the developing text and the content represented by the system.
Systems should provide some means for resolving these possible sources of conflict. The
existence of gaps in understanding need not be detrimental. Conflict resolution has been shown
to be an important part of group work; if a system can make conflict visible, and support the
negotiation and resolution of conflict, then it will more effectively encourage the development of
a common ground and shared understanding.
One way to reduce the possibility of these discontinuities developing in the different
shared spaces is to provide links between the spaces. In Calliope this is accomplished through
the use of annotations. This could be extended to allow other spaces, such as the scratchpad or a
common report space, to be linked to the main text. The system could also support any number
of “main” text spaces, all equally central to the discussion, and allow links to be made arbitrarily
between these spaces, extending the text to hypertext. The annotation mechanism could also be
extended to allow greater use of external references. Currently, Calliope allows the main text to
be linked to external text represented in the World Wide Web. The body of text being developed
by a group could be represented as a network of text fragments, or Web pages, any of which
could be edited at a given time by the various group members. This could act as a link between
synchronous and asynchronous collaborative writing.
In this thesis I have explored the impact of the use of collaborative writing software on
two groups of grade six students. Based on a case study of collaborative writing, I derived a
series of design requirements for synchronous collaborative writing tools, and built a prototype
to explore these requirements. The technology being used was specifically designed to assist
collaborative work, and the tasks assigned to the students were intended to make extensive use of
this technology. As this thesis has shown, the introduction of technology is not a neutral act, and
should not be seen as such. The design of computer systems to support collaborative writing
must take into consideration the effects of the system on social interaction as well as on the
performance of the task of writing together.
By extending my study over a period of three months, and then taking a detailed look at
the data from a multidisciplinary perspective, I was able to observe the wider ranging impacts of
the technology on group formation, interaction and control within the group, and development of
shared understanding. Qualitative field studies are an important means of investigating the use
of technology. The use of this approach to the study and development of collaborative
technology recognizes the impact of technology on the social fabric of reality, and incorporates
the act of designing technological artifacts into the interactive process of creating meaning in
society.
149
Appendix A: Notation for Transcripts
(adapted from (Goodman, 1981))
: sound preceding is noticeably prolonged- cut-off, abrupt terminationitalics emphasis (increased pitch or amplitude)
punct intonation. falling? rising
, falling/rising(1) seconds of pause( overlap begins
// overlap with next at new line(( simultaneous° lowered volume
< hurried start>> acceleration<< deceleration
= latch (new turn following previous with no pause)( ) unable to recover(word) guess
[comment] other actions, including: touch other, gaze at computer (own/others), touchcomputer (own/others), type, gesture, read from screen
I did not make use of the conventional notation for gaze (i.e. lines to show the start of focusingof gaze) since the timing of gaze shifts was not crucial for this analysis.
151
Appendix B: Samples from Data
(note that spelling and formatting are as in the students’ magazine)
Cover of the students’ magazine
(by Ryan, Wednesday group)
Selections from Wednesday group’s magazine section
Table of Contents
#1 Introduction*
#2 Perception by Different Group#3 When I’m in the Minority I Feel*#4 New Student#5 Prejudice Story*
#6 Word Find#7 What I Asked the Adviser#8 Clumsy Class
#9 Collage#10 What Happens Around the World* included below
152
Introduction
In this magazine our group has put together our thoughts and information on differenttopics that cover prejudice matters. We interviewed different people of different ages and
gathered their thoughts on prejudice acts. We decided to make this magazine because there areso many unsolved problems in the 90’s, and we hope that the people that read this magazine willget an idea of whats happening in the world when it comes to prejudist.
Minority poem
W h e n I ’ m i n t h eW h e n I ’ m i n t h em i n o r i t y I f e e l . . . . . .m i n o r i t y I f e e l . . . . . .
Very left out and I feel very differentbecause of one little difference. I sometimes feel disapointed.
For instance take a magority of people,that wanted to do one thing, and I wanted to do something different I would be in minority.
But I do feel good about myself if I know that it’s wrong.If its something good then I feel happy.I think a minority can have it’s ups and downs.
Another of our thoughts is on minoriy is, when a person is in a minority,
they feel in a way a little left out, because when you go to a different country,and don’t know how to speak the language,
it looks as though everyone is having so much fun,but you didn’t understand one word that was being said.You would be quite sad and lonley
Being in a minorty can be good and bad in many different ways.
-------By: Ryan Sarah Carla Kate--------
153
Prejudice Story by Kate, Ryan, and Carla
Tom was walking home from school, when he saw a scruffy old man talking to a wall.Tom looked at the man in great horror, and ran home.
The next day when he went to school, it was his turn for show and tell, he decided to talk
about the old man he saw, he started with saying that he saw a disgusting old man that chasedhim home and that started yelling at him. When he finished show and tell, he gathered all hisfriends and told them to meet him at the corner of Brunswick Ave.
When the bell rang, him and his friends ran to the corner, when they caught one glimpseof the mans face, and got a wiff of his breath they all started teasing him and laughing behind hisback. The man thought to himself, “Whats wrong with me? Nobody has the right to laugh at me
and call me names.” The man decided talk to the group of boys and girls. At first they werefrightened of him but then they realized what a kind and humourous man he was. The man feltvery good because he had just told a group of kids not the be prejudiced. The next day the kids
went to see the old man and before they new it they were having a good time talking andlaughing with him.
THE END
Selections from Tuesday group’s magazine section
Table of Contents
#1 Introduction*#2 Prejudice Story: Tiger Lily*#3 Letter to Jean Chretien
#4 Interview with League for Human Rights#5 Do’s and Don’t’s: New Student#6 Perception
#7 Girls Are Bad At Math#8 Girls Like Pink#9 When I’m in the minority*
#10 Clumsy Class#11 Similarities: Happy & Sad#12 Racism Survey* included below
154
Introduction
We decided to work on the problem of how to stop prejudice because we thought that itwas important for people to realize that prejudice hurts everyone. By reading this magazine wehope that you will realize that prejudice is a major problem that we need to solve, and hope that
you will understand more about this.Before we lead you into the contents of this magazine we would like to introduce
ourselves. Our names are David Lizoain, Elisa Morera, Faith Tomes and Samantha Yee. We are
grade five and six students and we all attend Huron Street Public School.
Minority poem
When I’m in the Minority I feel ...
Sometimes I feel so alone with nothing to do,Because all the people expect from me,More than what I can be.
I just can’t take it any longer,It’s not my fault,Though it’s not quite bad,
Who and what I am.But I like it sometimes when I’m the only one because peopleCan learn from me,
Things they never knew before,It’s sometimes good to be in the Minority,Because in the Majority people,
They might be doing bad things,I guess that I would feel different in different positions.
If I had 38 toes and 88 eyes I would feel VERY strange being,Near people with only 10 toes and 2 eyes,If I had wings and could fly,I might not be happy,
But I might not be sad.If I had no ears and no eyes,I would feel sad because I wouldn’t be able to see or hear,
But if I had the best marks in the world I would be happy as a clam!It really depends on which position I hold, and what purpose it served!
155
When I am in minority I feel bad when my idea doesn’t count,But if my idea is the one used or that works I feel good that Iwas the only one to come up with it.
Being in the minority just depends on what your position becauseif what you think doesn’t work you feel bad but if it works you feel good.
Sometimes it’s as if my opinion does not count,As if no one cares what I think.But sometimes when I’m in the Minority
I feel lucky,Different,One in a thousand.
Sometimes it’s fun being in the Minority,Even though it can be quite upsetting.
Prejudice Story by David, Elisa, Faith and Samantha
A Tale of Two Families
The year was 1956 and two families the Bagoodas and the Mohammeds lived next doorin Delhi, India. Both families were very poor and lived in very small huts. Mrs. Bagooda andMrs. Mohammed were both expecting babies. By coincidence on a hot dry day in June they both
gave birth to baby girls. When women are married in India they have to pay a dowry to thegrooms family. The Bagoodas did not want to pay the dowry when it came time for their child tobe married so they killed their baby daughter. However, the Mohammeds decided that they
would let their baby daughter live. They named her Tiger Lily which meant life.Mrs. Mohammed received lots of criticism from the other family because they thought
they should have also killed their daughter because the Mohammeds were also poor. Another
reason for this prejudice was because the Bagoodas were jealous that the Mohammeds had sucha beautiful, lively daughter who was always kind and friendly.
When Tiger Lily reached her teenage years a dreadful fire ravaged Delhi. The fire
quickly spread and soon reached the Bagoodas hut. When Tiger Lily was the Bagoodas hut wasin flames she rushed inside their hut even though they were not kind to her. After the entireBagooda family was outside of their hut she noticed Mrs. Bagooda was slumped unconscious on
the floor and was surrounded by flames. Tiger Lily quickly ran through the flames and rescuedMrs. Bagooda from a fiery death. Later Mrs. Bagooda apologized to the Mohammeds for herprejudism against them and lived the rest of her life wondering what her daughter would have
been like.
156
157
References
Allen, N.J., Atkinson, D., Morgan, M., Moore, T., and Snow, C. (1987) What ExperiencedCollaborators Say About Collaborative Writing. In Iowa State Journal of Business and
Technical Communication. September 1987, pp. 70-90.Apple Computer Inc. (1988a). Inside Macintosh Toolbox.Apple Computer Inc. (1988b). Macintosh Programmer’s Workshop C 3.0 Reference.
Baecker, R.M., Nastos, D., Posner, I.R., and Mawby, K.L. (1993). The User-centred IterativeDesign of Collaborative Writing Software. In Proceedings of InterCHI ‘93. New York:ACM Press, 399-405, 541.
Baecker, R.M., Glass, G., Mitchell, A., and Posner, I.R. (1994). SASSE: the CollaborativeEditor, 8 minute refereed video tape presented at the 1994 ACM Conference on HumanFactors in Computing Systems, May 1994; also published in SIGGRAPH Video Review,
1994.Beck, E. (1993). A Survey of Experiences of Collaborative Writing. In M. Sharples (ed.)
Computer Supported Collaborative Writing. New York: Springer-Verlag, pp. 87-112.
Beck, E., and Bellotti, V. (1993). Informed Opportunism as Strategy: Supporting Coordinationin Distributed Collaborative Writing. In Proceedings of the Third European Conferenceon Computer-Supported Cooperative Work - ECSCW ‘93. London: Kluwer Academic
Publishers, pp. 233-248.Buroway, M., Burton, A., Ferguson, A., Fox., K., Gamson, J., Gartrell, N., Hurst, L., Kurzman,
C., Salzinger, L., Schiffman, J., and Ui, S. (1991). Ethnography Unbound: Power and
Resistance in the Modern Metropolis. Berkeley: University of California Press.Buxton, B. (1992). Telepresence: Integrating shared task and person spaces. In Proceedings of
Graphics Interface ‘92. Toronto: Canadian Information Processing Society, pp. 123-129.
Carey, J.W. (1975). A Cultural Approach to Communications. In Communication. (2):1-22.Chalfonte, B.L., Fish, R.S., and Kraut, R.E. (1991). Expressive Richness: A Comparison of
Speech and Text as Media for Revision. In Proceedings of CHI’91: Conference on
Human Factors in Computing Systems. New York: ACM Press, pp. 21-26.Clark, H., and Schaefer, E. (1989). Contributing to Discourse. In Cognitive Science. 13(2), pp.
259-294.
Clark, H., and Brennan, S. (1991). Grounding in Communication. In L. Resnick, Levine, J., andTeasley, S. (ed.) Perspectives on Socially Shared Cognition. Washington, D.C.:American Psychological Society, pp. 127-149. Reprinted in Ronald Baecker (ed.)
Readings in Groupware and Computer-Supported Cooperative Work. Toronto: MorganKaufmann, pp. 222-234.
Cooper, M. (1982). Context as Vehicle: Implications in Writing. In Martin Nystrand (ed.) What
Writers Know: The Language, Process, and Structure of Written Discourse. Toronto:
158
Academic Press, pp. 105-128.
Csikszentmihalyi, M., and Rochberg-Halton, E. (1981). The Meaning of Things. Cambridge:Cambridge University Press.
Daiute, C., and Dalton, B. (1988). “Let’s Brighten It Up a Bit”: Collaboration and Cognition in
Writing. In B. Rafoth and Rubin, D. (ed.) The Social Construction of WrittenCommunication. Norwood, N.J.: Ablex Publishing Corporation, pp. 249-269.
Davies, G. (1989). Writing Without a Pencil. In M. Styles (ed.) Collaboration and Writing.
Milton Keynes: Open University Press, pp. 93-105.Derber, C. (1983). The Pursuit of Attention: Power and Individualism in Everyday Life. Toronto:
Oxford University Press.
Derycke, A. (1992). Towards a Hypermedium for Collaborative Learning? In Anthony R. Kaye(ed.) Collaborative Learning Through Computer Conferencing: The Najaden Papers.New York: Springer-Verlag, pp. 211-223.
Dhar, V., and Olson, M.H. (1989). Assumptions underlying systems that support work groupcollaboration. In M. Olson (ed.) Technological Support for Work Group Collaborations.Hilldale, N.J.: Lawrence Erlbaum Associates, pp. 33-50.
Dillon, A. (1993). How Collaborative is Collaborative Writing? An Analysis of the Productionof Two Technical Reports. In M. Sharples (ed.) Computer Supported CollaborativeWriting. New York: Springer-Verlag, pp. 69-85.
Dourish, P., and Bellotti, V. (1992). Awareness and Coordination in Shared Workspaces. InCSCW’92: Proceedings of the Conference on Computer-Supported CooperativeWork.Toronto: ACM Press, pp. 107-114.
Ede, L., and Lunsford, A. (1990). Singular Texts/Plural Authors: Perspectives on CollaborativeWriting. Carbondale, Illinois: Southern Illinois University Press.
Ellis, C.A., and Gibbs, S.J. (1989). Concurrency Control in Groupware Systems. In
Proceedings of ACM SIGMOD 1989. New York: ACM Press, pp. 343-355.Ellis, C.A., Gibbs, S.J., and Rein, G. (1989). Design and Use of a Group Editor. In Proceedings
of the IFIP Engineering for Human-Computer Interaction Conference. pp. 13-25.
Ellis, C.A., Gibbs, S.J., and Rein, G.L. (1991). Groupware: Some Issues and Experiences. InCommunications of the ACM. 34(1), pp. 38-58.
Fafchamps, D. (1991). Ethnographic Workflow Analysis: Specifications for Design. In
Bullinger, H.J. (ed.) Human Aspects in Computing: Design and Use of InteractiveSystems and Work with Terminals. Elsevier, pp. 709-715. Reprinted in Ronald Baecker(ed.) Readings in Groupware and Computer-Supported Cooperative Work. Toronto:Morgan Kaufmann, 1993, pp. 218-221.
Fitzgerald, J. (1992). Towards Knowledge in Writing: Illustrations from Revision Studies. NewYork: Springer-Verlag.
Flor, N.V., and Hutchins, E.L. (1991). Analyzing Distributed Cognition in Software Teams: A
Case Study of Team Programming During Perfective Software Maintenance. In J.
159
Koenemann-Belliveau, Moher, T., and Robertson, S. (ed.) Empirical Studies of
Programmers: Fourth Workshop. Norwood, N.J.: Ablex Publishing Corporation, pp. 36-63.
Flower, L.S., and Hayes, J.R. (1984). Images, Plans, and Prose: The Representation of Meaning
in Writing. In Written Communication. (1), pp. 120-160.Flower, L.S., Schriver, K.A., Carey, L., Haas, C., and Hayes, J.R. (1989). Planning in Writing:
The Cognition of a Constructive Process. Technical Report, Centre for the Study of
Writing, University of California, Berkeley.Freedman, S.W., Dyson, A.H., Flower, L., and Chafe,W. (1987). Research in Writing: Past,
Present and Future. Technical Report, Centre for the Study of Writing, University of
California, Berkeley.Fulk, J., Schmitz, J.A., and Schwarz, D. (1992). The Dynamics of Context-Behaviour
Interactions in Computer-Mediated Communication. In Martin Lea (ed.) Contexts of
Computer-Mediated Communication. New York: Harvester Wheatsheaf, pp. 7-29.Glaser, B., and Strauss, A. (1974). The Discovery of Grounded Theory: Strategies for Qualitative
Research. Chicago: Aldine.
Goldman-Segall, R. (1991). A Multimedia Research Tool for Ethnographic Investigation. In I.Harel and Papert, S. (ed.) Constructionism. Norwood, N.J.: Ablex PublishingCorporation, pp. 467-497.
Goodman, C. (1981). Conversational Organization: Interaction Between Speakers and Hearers.New York: Academic Press.
Greenberg, S., and Bohnet, R. (1991). GroupSketch: A Multi-User Sketchpad for Geographically
Distributed Small Groups. In Proceedings of Graphics Interface ‘91. Calgary, Alberta,pp. 207-215.
Greenberg, S., and Marwood, D. (1994). Real Time Groupware as a Distributed System:
Concurrency Control and Its Effect on the Interface. In CSCW’94: Proceedings of theConference on Computer-Supported Cooperative Work. New York: ACM Press, pp. 207-217.
Greif, I., Seliger, R., and Weihl, W. (1992). A Case Study of CES: A Distributed CollaborativeEditing System Implemented in Argus. In IEEE Transactions on Software Engineering.18(9), pp. 827-839..
Group Technologies Inc. (1990). Aspects: Simultaneous Conference Software for the Macintosh.Grudin, J. (1989). Groupware and Cooperative Work: Problems and Prospects. In Brenda Laurel
(ed.) The Art of Human-Computer Interface Design. Reading, Mass.: Addison-WesleyPublishing Company, pp. 171-185.
Gubrium, J.F. (1988). Analyzing Field Reality. Newbury Park, CA: Sage Publications.Gutwin, C., Stark, G., and Greenberg, S. (1995). Support for Workspace Awareness in
Educational Groupware. In CSCL ‘95: Computer Support for Collaborative Learning.
Mahwah, N.J.: Lawrence Erlbaum Associates, pp. 147-156.
160
Harrison, B., Owen, R., and Baecker, R.M. (1994). Timelines: An Interactive System for the
Collection and Visualization of Temporal Data. In Proceedings of Graphical Interface‘94. Toronto: Canadian Information Processing Society, pp. 141-148.
Hayes, J.R., and Flower, L.S. (1980). Identifying the Organization of Writing Processes. In
Gregg and Steinberg (ed.)Cognitive Processes in Writing. Hilldale, N.J.: LawrenceErlbaum Associates, pp. 3-30.
Hughes, P.T. (1993). Going Off the Rails: Understanding Conflict in Practice. In S. Easterbrook
(ed.) CSCW: Cooperation or Conflict. Berlin: Springer-Verlag, pp. 161-170.Ishii, H., Minoru, K., and Grudin, J. (1992). Integration of Inter-Personal Space and Shared
Workspace: ClearBoard Design and Experiments. In J. Turner and Kraut, R. (ed.)
Proceedings of CSCW ‘92: Conference on Computer-Supported Cooperative Work.Toronto: ACM Press, pp. 33-43.
Karimi, S., and Palevich, J. (1990). Interact: A User Study; A Real-time Shared Document
Authoring and Editing System for Collaboration. Technical Report, AdvancedTechnology Group - Human Interface Group, Apple Computer, June 1990.
Karsenty, A., and Beaudouin-Lafon, M. (1993). An Algorithm for Distributed Groupware
Applications. In The 13th Conference on Distributed Computing Systems. New York:IEEE Press, pp. 195-202.
Kaye, A.R. (1993). Computer Networking for Development of Distance Education Courses. In
M. Sharples (ed.) Computer Supported Collaborative Writing. New York: SpringerVerlag, pp. 41-67.
Kraut, R.E., Egido, C., and Galegher, J. (1990). Patterns of Contact and Communication in
Scientific Research Collaborations. In J. Galegher Kraut., R., and Egido, C. (ed.)Intellectual Teamwork: Social and Technological Foundations of Cooperative Work.Hillsdale, N.J.: Lawrence Erlbaum Associates, pp. 149-171.
Lauwers, J.C., and Lantz, K.A. (1990). Collaboration Awareness in Support of CollaborationTransparency: Requirements for the Next Generation of Shared Window Systems.Proceedings of CHI ‘90: Conference on Human Factors in Computing Systems. New
York: ACM Press. pp. 303-311.Lea, M. (1991). Rationalist assumptions in cross-media comparisons of computer mediated
communication. In Behaviour and Information Technology. 10(2), pp. 153-172.
Leeds-Hurwitz, W. (1993). Semiotics and Communication: Signs, Codes, Cultures. Hillsdale,N.J.: Lawrence Erlbaum Associates.
Leland, M.D.P., Fish, R.S., and Kraut, R.E. (1988). Collaborative Document Production UsingQuilt. CSCW ‘88: Proceedings of the Conference on Computer-Supported Cooperative
Work. New York: ACM Press, pp. 206-215.Mantei, M.M. (1988). Capturing the Capture Lab Concepts: A Case Study in the Design of
Computer Supported Meeting Environments. Proceedings of CHI ‘88: Conference on
Human Factors in Computing Systems. Portland, OR.: ACM Press, pp. 257-270.
161
Mason, R. (1992). Evaluation Methodologies for Computer Conferencing Applications. In
Anthony R. Kaye (ed.) Collaborative Learning Through Computer Conferencing: TheNajaden Papers. New York: Springer-Verlag, pp. 105-116.
Mawby, K.L. (1991). Designing Collaborative Writing Tools. M.Sc. Thesis. Department of
Computer Science, University of Toronto.McCarthy, J., Miles, V. and Monk, A. (1991). An Experimental Study of Common Ground in
Text-Based Communication. Proceedings of CHI’91: Conference on Human Factors in
Computing Systems. New York: ACM Press, pp. 209-215.McGrath, J. (1993). Groups: Interaction and Performance. Englewood Cliffs, N.J.: Prentice
Hall. Exerpted in Ronald Baecker (ed.) Readings in Groupware and Computer-
Supported Cooperative Work. Toronto: Morgan Kaufman, pp. 200-205.McGuffin, L., and Olson, G.M. (1992). ShrEdit, a Shared Electronic Workspace. Cognitive
Science and Machine Intelligence Laboratory, University of Michigan.
McLaughlin Hymes, C. and Olson, G.M. (1992). Unblocking Brainstorming Through the Use ofa Simple Group Editor. CSCW ‘92: Proceedings of the Conference on Computer-Supported Cooperative Work. Toronto: ACM Press, pp. 99-106.
McMahon, H., and O’Neill, W. (1992). Computer-Mediated Zones of Engagement. In T. Duffy,Lowyck, J., Jonassen, D., and Welsh, T. (ed.) Designing Environments for ConstructiveLearning. New York: Springer-Verlag, pp. 37-57.
MetroWerks (1996). Inside MetroWerks CodeWarrior C/C++.Miles, V.C., McCarthy, J.C., Dix, A.J., Harrison, M.D., and Monk, A.F. (1993). Reviewing
Designs for a Synchronous-Asynchronous Group Editing Environment. In M. Sharples
(ed.) Computer Supported Collaborative Writing. New York: Springer Verlag, pp. 137-160.
Minör, S., and Magnuson, B. (1993). A Model for Semi-(a)Synchronous Collaborative Editing.
In Proceedings of the Third European Conference on Computer-Supported CooperativeWork - ECSCW ‘93. London: Kluwer Academic Publishers, pp. 219-231.
Nastos, D. (1992). A Structured Environment for Collaborative Writing. M.Sc. Thesis.
Department of Computer Science, University of Toronto.Neeri, M. (1996). Grand Unified Socket Interface.Neill, S.D. (1993). Clarifying McLuhan: An Assessment of Process and Product. Wesport,
Conneticut: Greenwood Press.Neuwirth, C.M., and Kaufer, D.S. (1989). The Role of External Representations in the Writing
Process. In Proceedings of Hypertext ‘89. pp. 319-342.Neuwirth, C.M., Kaufer, D.S., Chandhok, R., and Morris, J.H. (1990). Issues in the Design of
Computer Support for Co-authoring and Commenting. In CSCW ‘90: Proceedings of theConference on Computer-Supported Cooperative Work. New York: ACM Press, pp. 183-195.
Neuwirth, C.M., Chandhok, R., Kaufer, D.S., Erion, P., Morris, J., and Miller, D. (1992).
162
Flexible DIFF-ing in a Collaborative Writing System. In CSCW ‘92: Proceedings of the
Conference on Computer-Supported Cooperative Work. Toronto: ACM Press, pp. 147-154.
Neuwirth, C.M., Kaufer, D., Chandhok, R., and Morris, J. (1994). Computer Support for
Distributed Collaborative Writing: Defining Parameters of Interaction. CSCW ‘94:Proceeddings of the Conference on Computer-Supported Cooperative Work. pp. 145-152.
Newman, J., and Newman, R. (1992). Three Modes of Collaborative Authoring. In P. Holt and
Williams, N. (ed.) Computers and Writing: State of the Art. London: Kluwer AcademicPublishers, pp. 20-28.
Nystrand, M. (1982). The Structure of Textual Space. In Martin Nystrand (ed.) What Writers
Know: The Language, Process, and Structure of Written Discourse. Toronto: AcademicPress, pp. 75-86.
Nystrand, M. (1986). The Structure of Written Communication: Studies in Reciprocity between
Writers and Readers. Toronto: Academic Press.Nystrand, M. (1989). A Social-Interactive Model of writing. In Written Communication. (6),
pp. 66-85.
Nystrand, M., and Brandt, D. (1989). Response to Writing as a Context for Learning to Write. InC. Anson (ed.) Writing and Response: Theory, Practice, and Research. Urbana, IL:National Council of Teachers of English, pp. 209-230.
Olson, J.S., Olson, G.M., Mack, L.A., and Wellner, P. (1990). Concurrent Editing: The Group’sInterface. In Proceedings of Interact ‘90. pp. 835-840.
Olson, J.S., Olson, G.M., Storrøsten, M., and Carter, M. (1992). How a Group-Editor Changes
the Character of a Design Meeting as well as its Outcome. CSCW ‘92 Proceedings of theConference on Computer-Supported Cooperative Work. New York: ACM Press, pp. 91-98.
Olson, J.S., Olson, G., and Meader, D. (1995). What Mix of Video and Audio is Useful for SmallGroups Doing Remote Real-Time Work? In Proceedings of CHI ‘95: Conference onHuman Factors in Computing Systems. New York: ACM Press, pp. 362-369.
Osterhout, J. (1994). Tcl and the Tk Toolkit. Toronto: Addison Wesley.Pacull, F., Sandoz, A., and Schiper, A. (1994). Duplex: A Distributed Collaborative Editing
Environment in Large Scale. In CSCW ‘94: Proceedings of the Conference on Computer-
Supported Cooperative Work. New York: ACM Press, pp. 165-173.Pontecorvo, C. (1992). Developing Literacy Skills Through Cooperative Computer Use: Issues
for Learning and Instruction. Designing Environments for Constructive Learning. NewYork: Springer-Verlag, pp. 139-160.
Posner, I.R. (1991). A Study of Collaborative Writing. M.Sc. Thesis. Department of ComputerScience, University of Toronto,
Posner, I.R., and Baecker, R.M. (1992). How People Write Together. The Twenty-fifth Annual
Hawaii International Conference on System Sciences. pp. 127-138. Reprinted, with
163
slight modification, in Ronald Baecker (ed.) Readings in Groupware and Computer-
Supported Cooperative Work.. Toronto: Morgan Kauffman, 1993, pp. 127-138.Posner, I.R., Mitchell, A., and Baecker, R.M. (1996 (in press)). Learning to Write Together
Using Groupware. In R. Rada (ed.) Groupware and Authoring. London: Academic
Press. Revised version of Mitchell, A., Posner, I.R., and Baecker, R.M., Learning toWrite Together Using Groupware. In Proceedings of CHI ‘95: Conference on HumanFactors in Computing Systems. New York: ACM Press, pp. 288-295.
Rafoth, B. (1988). Discourse Community: Where Writers, Readers, and Texts Come Together.In B. Rafoth and Rubin, D. (ed.) The Social Construction of Written Communication.Norwood, N.J.: Ablex Publishing Corporation, pp. 131-146.
Reddy, M. (1979). The Conduit Metaphor - a Case of Frame Conflict in Our Language AboutLanguage. In A. Ortony (ed.) Metaphor and Thought. Cambridge, U.K.: CambridgeUniversity Press, pp. 284-324.
Resnick, L. (1991). Shared Cognition: Thinking as Social Practice. In L. Resnick, Levine, J.,and Teasley, S. (ed.) Perspectives on Socially Shared Cognition. Washington, D.C.:American Psychological Society, pp. 1-20.
Rohman, G. (1965). Pre-writing: The Stage of Discovery in the Writing Process. In CollegeComposition and Communication. (29), pp. 209-211.
Roseman, M. (1993). Design of a Real-Time Groupware Toolkit. M.Sc. Thesis. Department of
Computer Science, University of Calgary.Roseman, M., and Greenberg, S. (1996). Building Real Time Groupware with GroupKit, A
Groupware Toolkit. In ACM Transactions on Computer Human Interaction (TOCHI).
Accepted for publication July 1995.Rubin, D. (1988). Introduction: Four Dimensions of Social Construction in Written
Communication. In B. Rafoth and Rubin, D. (ed.) The Social Construction of Written
Communication. Norwood, N.J.: Ablex Publishing Corporation, pp. 1-33.Scardamalia, M., Bereiter, C., and Goelman, H. (1982). The Role of Production Factors in
Writing Ability. In Martin Nystrand (ed.) What Writers Know: The Language, Process,
and Structure of Written Discourse. Toronto: Academic Press, pp. 173-210.Scardamalia, M., and Bereiter, C. (1992). An Architecture for Collaborative Knowledge-
Building. In Erik De Corte Marcia C. Linn, Heinz Mandl and Lieven Verschaffel (ed.)
Computer-Based Learning Environments and Problem Solving. New York: Springer-Verlag, pp. 229-248.
Shapiro, D. (1994). The Limits of Ethnography: Combining Social Sciences for CSCW. CSCW‘94: Proceedings of the Conference on Computer-Supported Cooperative Work: New
York: ACM Press, pp. 417-428.Sharples, M., and Pemberton, L. (1992). Representing Writing: External Representations and the
Writing Process. In P. Holt and Williams, N. (ed.) Computers and Writing: State of the
Art.. London: Kluwer Academic Publishers, pp. 319-336.
164
Sharples, M., Goodlet, J.S., Beck, E.E., Wood, C.C., Easterbrook, S.M., and Plowman, L.
(1993). Research Issues in the Study of Computer Supported Collaborative Writing. InM. Sharples (ed.) Computer Supported Collaborative Writing. New York: SpringerVerlag, pp. 9-28.
Silberschatz, A., and Galvin, P. (1994). Operating Systems Concepts. Don Mills, Ontario:Addison-Wesley.
Stefik, M., Foster, G., Bobrow, D.G., Kahn, K., Lanning, S., and Suchman, L. (1987). Beyond
the Chalkboard: Computer Support for Collaboration and Problem Solving in Meetings.In Communications of the ACM. 30(1), pp. 32-47.
Suchman, L. (1987). Plans and Situated Actions. New York: Cambridge University Press.
Suchman, L., and Trigg, R.H. (1991). Understanding Practice: Video as a Medium for Reflectionand Design. In J. Greenbaum and Kyng, M. (ed.) Design at Work: Cooperative Design ofComputer Systems. Hillsdale, N.J.: Lawrence Erlbaum, pp. 65-89. Reprinted in Ronald
Baecker (ed.) Readings in Groupware and Computer-Supported Cooperative Work..Toronto: Morgan Kauffman, 1993, pp. 205-217.
Tatar, D., Foster, G., and Bobrow, D. (1991). Design for Conversation: Lessons from Cognoter.
In International Journal of Man-Machine Studies. (34), pp. 185-209.Turkle, S. (1984). The Second Self: Computers and the Human Spirit. New York: Simon and
Schuster Inc.
Vidich, A., and Lyman, S. (1994). Qualitative Methods: Their History in Sociology andAnthropology. In N. Denzin and Lincoln, Y. (ed.) Handbook of Qualitative Research..London: Sage Publications, pp. 23-59.
Vygotsky, L. (1978). Mind in Society. Cambridge, MA: Harvard University Press.Wagonner, M. (1992). A Case Study Approach to Evaluation of Computer Conferencing. In
Anthony R. Kaye (ed.) Collaborative Learning Through Computer Conferencing: The
Najaden Papers. New York: Springer-Verlag, pp. 137-146.Zhang, J. (1994). Representations in Distributed Cognitive Tasks. In Cognitive Science. (18), pp.
87-122.