tying knowledge to action with kmail - biuteenid/publications/kmail.pdfthe kmail par adigm g rew...

7
KNOWLEDGE MANAGEMENT Tying Knowledge to Action with kMail David G. Schwartz and Dov Te’eni, Bar-Ilan University MAY/JUNE 2000 1094-7167/00/$10.00 © 2000 IEEE 33 Y OU’VE GOT THIS MASSIVE, ALL- encompassing organizational memory. Now what? To make effective use of the OM’s knowledge, you’ve got to tie it to action. But tying knowledge to action can only occur if you are aware of the knowledge and can identify it at the time of action and if the sys- tem can deliver it to that point of action. For that, you need tools. Knowledge-enhanced e-mail, kMail for short, is just such a tool. This new class of e-mail application intertwines e-mail with knowledge management, thus representing a fundamental shift in the way we can distrib- ute and maintain organizational knowledge. In this article, we emphasize knowledge distribution and tie it into organizational action. Knowledge links up with action by a process of contextualization. To make these ideas work in practice, we rely on the Inter- net and e-mail as a transport layer for knowl- edge dissemination. We also describe the core components and techniques that help decide when to contextualize messages and with what knowledge. Knowledge management: beyond time and space The knowledge of man is as the waters, some descending from above, and some springing from beneath. —Francis Bacon (1561–1626) Knowledge management in general, and OMs in particular, are an attempt by the orga- nization to transcend time and space in learn- ing. In building knowledge-management sys- tems (KMSs), we attempt to assimilate knowledge available within the organization and disseminate it to people connected to the organization. Challenges associated with knowledge management fall into three general categories: acquisition, organization, and distribution. Knowledge acquisition deals with the issues that surround knowledge extraction in its various forms—from the organization’s knowledge bases, databases, printed re- sources, and people. In knowledge acquisi- tion, we must deal with questions such as “Who is the authority on a specific area?” “How is knowledge in that area currently stored?” and “How can I get my hands on the knowledge and make it machine readable?” Knowledge organization deals with the issues surrounding how to best store knowl- edge so that it can be retrieved when relevant. In knowledge organization, we must deal with questions such as “In what forms should we store this knowledge?” “How should we index this knowledge?” and “How is the user going to ask for this knowledge?” Knowledge distribution, or dissemination, must tackle the problem of getting the right knowledge to the right place at the right time. It requires three conditions: Awareness: The user must be aware that there is relevant knowledge available. Does the manager know that a best-prac- tices database exists? Does he or she know that there may be something in there to help resolve the current crisis? THE KMAIL SYSTEM INTEGRATES E-MAIL WITH ORGANIZATIONAL MEMORIES TO DELIVER KNOWLEDGE ACROSS THE INTERNET IN A TIMELY , RELEVANT MANNER. KMAIL OPERATES BY PROVIDING CONTEXT THROUGH METAKNOWLEDGE-BASED MEMORYCONCEPT ASSOCIATIONS AND BY DETERMINING THE APPROPRIATE MEMORY ITEMS THROUGH THE CREATION OF OM VIEWS.

Upload: others

Post on 26-Sep-2020

0 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Tying Knowledge to Action with kMail - BIUteenid/publications/kmail.pdfThe kMail par adigm g rew from a resear ch project we undertook to investigate ways to improve intercultural

K N O W L E D G E M A N A G E M E N T

Tying Knowledge to Actionwith kMail

David G. Schwartz and Dov Te’eni, Bar-Ilan University

MAY/JUNE 2000 1094-7167/00/$10.00 © 2000 IEEE 33

YOU’VE GOT THIS MASSIVE, ALL-encompassing organizational memory. Nowwhat? To make effective use of the OM’sknowledge, you’ve got to tie it to action. Buttying knowledge to action can only occur ifyou are aware of the knowledge and canidentify it at the time of action and if the sys-tem can deliver it to that point of action. Forthat, you need tools.

Knowledge-enhanced e-mail, kMail forshort, is just such a tool. This new class ofe-mail application intertwines e-mail withknowledge management, thus representing afundamental shift in the way we can distrib-ute and maintain organizational knowledge.

In this article, we emphasize knowledgedistribution and tie it into organizationalaction. Knowledge links up with action by aprocess of contextualization. To make theseideas work in practice, we rely on the Inter-net and e-mail as a transport layer for knowl-edge dissemination. We also describe thecore components and techniques that helpdecide when to contextualize messages andwith what knowledge.

Knowledge management:beyond time and space

The knowledge of man is as the waters, somedescending from above, and some springingfrom beneath.—Francis Bacon (1561–1626)

Knowledge management in general, andOMs in particular, are an attempt by the orga-nization to transcend time and space in learn-ing. In building knowledge-management sys-tems (KMSs), we attempt to assimilateknowledge available within the organizationand disseminate it to people connected to theorganization.

Challenges associated with knowledgemanagement fall into three general categories:acquisition, organization, and distribution.

Knowledge acquisition deals with theissues that surround knowledge extraction inits various forms—from the organization’sknowledge bases, databases, printed re-sources, and people. In knowledge acquisi-tion, we must deal with questions such as“Who is the authority on a specific area?”“How is knowledge in that area currentlystored?” and “How can I get my hands on the

knowledge and make it machine readable?”Knowledge organization deals with the

issues surrounding how to best store knowl-edge so that it can be retrieved when relevant.In knowledge organization, we must dealwith questions such as “In what forms shouldwe store this knowledge?” “How should weindex this knowledge?” and “How is the usergoing to ask for this knowledge?”

Knowledge distribution, or dissemination,must tackle the problem of getting the rightknowledge to the right place at the right time.It requires three conditions:

• Awareness: The user must be aware thatthere is relevant knowledge available.Does the manager know that a best-prac-tices database exists? Does he or sheknow that there may be something inthere to help resolve the current crisis?

THE KMAIL SYSTEM INTEGRATES E-MAIL WITH

ORGANIZATIONAL MEMORIES TO DELIVER KNOWLEDGE ACROSS

THE INTERNET IN A TIMELY, RELEVANT MANNER. KMAIL

OPERATES BY PROVIDING CONTEXT THROUGH

METAKNOWLEDGE-BASED MEMORY–CONCEPT ASSOCIATIONS

AND BY DETERMINING THE APPROPRIATE MEMORY ITEMS

THROUGH THE CREATION OF OM VIEWS.

Page 2: Tying Knowledge to Action with kMail - BIUteenid/publications/kmail.pdfThe kMail par adigm g rew from a resear ch project we undertook to investigate ways to improve intercultural

• Identification: The user must be able toreadily identify that knowledge. Can themanager effectively use the best-practicesdatabase to find something that will helphim or her?

• Delivery:The knowledge must be deliveredto the point of need, in a timely manner.

Managers in an organization do not havethe time or inclination to actively seek orga-nizational knowledge. It would be far moreeffective if the knowledge could find them.

Consider the use of software help systems.No one reads through the help files of a soft-ware system. The biggest advance in help sys-tems of the past decade has been context-sen-sitive help,which is successful because it tiesknowledge to action. The use of organiza-tional knowledge must go through the sametransition.

We are enabling knowledge distributionfor action. The development of kMail is adirect response to the need to tie knowledgeto action,after considering the awareness,identification,and delivery aspects of knowl-edge distribution.

Moving knowledge management to theInter net. There are three major classes ofapplication dominating the Internet today, allof which are relevant to knowledge-enhancede-mail:

• Web sites that store inordinate amountsof occasionally useful information;

• Pop–client or browser-based e-mail ser-vices; and

• Web sites that provide extensive index-ing, cataloging, and search services.

Put the three of them together in an organi-zational setting and a transformation occurs:

• The Web site becomes an HTML-basedknowledge repository that can be viewedthrough an organizational lens.

• The e-mail service becomes a transparentinterface to manage knowledge distribution.

• The indexing and search incorporatesorganizational metaknowledge and userprofiles.

The end result is a kMail system.The kMail paradigm grew from a research

project we undertook to investigate ways toimprove intercultural communication inmultinational corporations.1,2 That research,resulting in the Hypermail system,showed

that there is a well-defined set of threats tocommunication. Our investigations found thatthere are two fundamental influences on theeffectiveness of electronic communication:cultural and organizational norms andsender–receiver distance, the latter dependingon personal background, experience, com-pany role, and other context-forming criteria.

The starting point of the communicationprocess is the goals of organizational com-munication. Based on Jurgen Habermas’swork,3 the categories of such goals would be

1. Commanding a specific action;2. Managing collective action:thinking

collectively; monitoring communica-tion,command, and control; and settingwork procedures and rules;

3. Influencing (persuasion, leadership,lobbying, propositions);

4. Providing information for future action(knowledge dissemination); and

5. Seeking information for future action(knowledge acquisition);

We can make each of these action-orientedgoals more attainable by tighter integrationof knowledge.

Why e-mail? When an organization’s man-agers communicate by e-mail,in most cases,they are trying to achieve some action orresponding to a request for some action.Those actions’effectiveness depends largelyon the correct understanding of what the bestaction should be. This, in turn, depends onknowing the proper context in which thataction is requested. The five goals of com-munication we’ve listed are equally valid forelectronic communication.

Daniel O’Leary4 discusses three signifi-

cant limitations to knowledge managementinitiatives that focus on knowledge as an endin itself.

• Knowledge does not necessarily result inaction to create value: organizationalknowledge-management initiatives mustbe action-oriented and create value if theyare to acquire the management priorityand heavy resources needed to fund theirdevelopment.

• Knowledge processes are dynamic:KMSs all suffer from one significant bar-rier—creating the knowledge and keep-ing it current.

• Knowledge cannot be a surrogate for cre-ativity: knowledge should inform andguide creativity. For this to happen,knowledge must be available during thecreative processes followed in day-to-daybusiness activities.

The question we posed to ourselves was“How can we best keep knowledge dynamic,use it in action-oriented situations,and makeit the backdrop for creativity?” The answeris through e-mail,the quintessential Internetapplication. Consider the following:

• Every organization, without exception,will have an e-mail infrastructure beforeit reaches the stage of developing an OM.

• E-mail communication in a modern orga-nization is over 78% action-oriented,according to a recent study.2 Organiza-tions must converge to action,and com-munication is perhaps the foundation formost organizational action.5

• Managers,and knowledge workers of allkinds,interact with their e-mail systemson a daily basis—it is a standard operat-ing procedure. This means that using e-mail as the window into an OM gives usthe smallest delta for change in an orga-nization’s daily activities.

• Managers are motivated to achieve suc-cessful communication. They want theirinstructions understood and their answersto queries to be effective.

There are many ways to connect people toknowledge through the use of intelligentagents,push technology, or search engines.But connecting people with knowledgeshould be transparent. A knowledge-man-agement tool is nearly worthless if managersmust learn specialized techniques or com-plex interfaces to get at that knowledge. Best-

34 IEEE INTELLIGENT SYSTEMS

MANAGERS IN AN ORGANIZATION

DO NOT HAVE THE TIME OR

INCLINATION TO ACTIVELY SEEK

ORGANIZATIONAL KNOWLEDGE.IT WOULD BE FAR MORE

EFFECTIVE IF THE KNOWLEDGE

COULD FIND THEM.

Page 3: Tying Knowledge to Action with kMail - BIUteenid/publications/kmail.pdfThe kMail par adigm g rew from a resear ch project we undertook to investigate ways to improve intercultural

practices database usage for every new caseis a far cry from the ubiquitous use of orga-nizational knowledge that businesses arelooking for.

Contextualization and organizationalknowledge. To use knowledge in an action-oriented manner, the e-mail messages mustbe able to access that knowledge in its correctcontext. Contextualization must rely on orga-nizational knowledge for two components:knowledge to provide the additional contextlayers around action and knowledge to iden-tify the conditions in which to contextualizemessages. We assume here that organiza-tional knowledge is explicit and available,which is extremely naive and outright wrongin many cases. However, new technologiessuch as expert support systems,intranets,anddocument (including e-mail document) man-agement are making it feasible to storeknowledge in an accessible fashion. Ourstarting point,therefore, assumes an existingURL-based OM,which could enable linkingthe knowledge to ongoing communication.

Core components

Retrieving knowledge in context is not atrivial task. To do so,kMail combines corecomponents—message parsing to identifyconcepts in an e-mail message; search and indexing to create the metaknowledgeoverview of the OM; and profile matching andlinking concepts—to memory items in theOM to minimize sender–receiver distance.

Accessing the OM through metaknowledge.Metaknowledge plays an essential role inconnecting e-mail with OMs. The identifi-cation of the sender and recipient of an e-mailmessage is of paramount importance. Know-ing this lets us access a wealth of user-spe-cific metaknowledge for use in focusingaccess to the OM. Of course, both the senderand recipient identifications are readily avail-able in an e-mail application.

The metaknowledge kMail uses consistsof two main components—user profile infor-mation and explicit memory–concept asso-ciation (MCA) information. Both reside in ahighly structured relational database. The(formal) metaknowledge serves as the linkbetween the (informal) e-mail communica-tions and the (semiformal) HTML-basedorganizational knowledge base. Figure 1shows a metalevel architecture.1,6

Structur ing the metaknowledge withMCAs. The OM metaknowledge relies on theuse of MCAs. An MCA exists when there iseither an explicit or derived association betweena given concept and a memory item in the OM.A concept can exist without being associatedwith a memory item,and a memory item canexist without being associated with a concept.

For example, we may associate the concept“chocolate”with any number of memory items:chocolate cookies,chocolate milk,or a choco-late color swatch. These associations range instrength from user-specific, to project-specific,to department-specific,and on to organization-specific associations. Determining the correctassociation between general concepts and spe-cific memory items is a function of OM views.

Contextualization thr ough metaknowledge.Mark Ackerman7 and others have addressedthe use of contextualization in organizationalmemories. Ackerman refers to context infor-mation about the specifics of the sender–receiver attributes as well as the task char-acteristics prevailing at the time of theknowledge generation. We call this type ofcontext information situational context. Heargues that contextual information mustoften be dropped when building an OM togeneralize the information. In establishingthe “correct” level of context, we wouldargue that although context can be removedfrom OMs,it should not be discarded. Onthe contrary, by distilling contextual infor-mation and retaining it in the form of meta-knowledge, we can achieve Ackerman’s goalof generalizable OMs,without losing ourability to recontextualize that knowledgewhen the application so demands. kMail isjust such an application. Having removedcontext to add longevity to the knowledge,we must then recontextualize that knowl-edge to add meaning to communications andtie knowledge to action.

Examples of situational context regardinga given item in the OM include attributessuch as

• time or date of creation,• time or date of modification,• name of author,• title of author,• current project of author,• number of years author is with the

company,• number of years author is in current

position,• department of author, and• department details (country or mandate,

for example).

Thus many context-free memory items canexist in the OM,only to have the context rein-troduced when a user needs to link a memoryitem with a specific call to action. Thisapproach is consistent with that of Ackermanand others who contend that if we wantknowledge to be a useful long-term resource,we should distill it into a context-free form.The context itself is a part of the metaknowl-edge, and we reintroduce it to the OM whenwe use that knowledge to perform an action.

OM views

Beauty might be in the eye of the beholder,but meaning is in the eye of the e-mail author.If we want to properly bring an OM to bearon the concepts used in a message, we mustfocus on the author’s intention to use thoseconcepts. This could only be possible if weview the OM through the author’s eyes.Although the degree of context requireddepends on the e-mail recipient,8 it is theauthor who must determine the context thatthe message provides,as well as its depth.There is a limit to the recipient’s input,at

MAY/JUNE 2000 35

User profiles

Organizational memory

Explicit memory-concept association

To:From:Re:

(a) (b) (c)

Form

al

Info

rmal

Figure 1. Three-tier architecture tying e-mail to organizational memory through metaknowledge: (a) applicationlevel—with concepts; (b) metalevel—formal metaknowledge; and (c) object level—semi-formal memory items.

Page 4: Tying Knowledge to Action with kMail - BIUteenid/publications/kmail.pdfThe kMail par adigm g rew from a resear ch project we undertook to investigate ways to improve intercultural

this stage, for controlling whether to accessthe contextual knowledge or just let it slideby transparently. In this respect,kMail beginsto address Buckingham Shum’s context para-dox,8 with different views provided to dif-ferent recipients and with the embeddedlinks’ nonintrusive nature letting the infor-mation overload regulate itself.

An OM view is similar in concept to adatabase view.9 Where a database view takesa database structure and produces a logicallyfiltered view suited to a given query, an OMview takes an OM’s logical content and pro-duces a filtered view suited to a given user ina certain situation. In both cases,the under-lying database or OM does not change—justthe way we see it. As with a database view,an OM view can be transient and has nophysical existence.

Creating views on the fly. To implementkMail, each time a user authors an e-mailmessage, the system creates a new OM view.The view consists of OM concepts relevant tothe concepts from the current e-mail,withinthe context of this user’s activities. The sys-tem first presents the author with this viewfor the purpose of confirmation, validation,

or modification and then sends the validatedview to the kMail recipient along with thekMail message.

In fact,there is no significant data trans-mission required on top of a standard e-mailmessage. Segments of the OM view link upwith the selected concepts in the kMail mes-sage. The OM itself stays in place. Only therelevant portion of the view travels across theInternet to the kMail recipient.

OM views for external consumption. Userscan also tailor OM views in kMail for exter-nal consumption. An increasingly attractivescenario for the use of KMSs involves open-ing up part of that knowledge to customersor strategic partners. This raises a host ofquestions regarding access and security. OMviews facilitate this type of activity. Thesender can administer a secure OM-view cre-ation mechanism in much the same way as adatabase scheme to determine what views heor she can make available to the outsideworld.

Creating OM views to minimize sender–receiver distance. To create an OM view andminimize sender–receiver distance, we per-

form a series of queries on the OM meta-knowledge database. These queries result informing the initial view presented to thesender.

Four views form the basis of the initialanalysis that the kMail server (Figure 2) per-forms on the OM,each view being tested bya different query on the OM metaknowledge.The resulting user view concatenates andcombines these queries. Alternatively, thesystem can present each view separately, aswith the departmental view in our examplebelow.

A series of tables represent the user pro-files and explicit MCAs in the metaknowl-edge. The key metaknowledge that we musteventually arrive at consists of the entries inthe memories table that best match the mem-ory item’s intended use. The memories table,however, is a context-free representation ofthe OM. It contains no indication of who hasused this memory item and for what purpose.Its purpose is to provide an unbiased pointerinto the collective OM that can then bemapped for appropriate use.

The memories table consists of four fields:

• MemoryID: a unique identifier automat-ically assigned to each new OM entry,

• MemoryURL: the URL where the mem-ory item can be found,

• MemoryDescription: a brief descriptionof the memory item,and

• MemoryType:a category code that indi-cates the type of contents in the memoryitem.

MemoryType is a significant piece ofmetaknowledge in that it lets us rank and sortOM entries based on intended usage. Table1 lists MemoryType codes and their corre-sponding meanings.

Other metaknowledge tables required toimplement the situational context attributesdiscussed earlier include

• People(PersonID, E-mail, Surname,Firstname),

36 IEEE INTELLIGENT SYSTEMS

OMDBOM

DBOMDB

OMDBOM

DBOMDB

OMDBOM

DB

kMail sender client1. Message parsing and concept identification3. Link insertion4. Link validation

OMDB

UserprofileMCAs

UserprofileMCAs OM

index

5. kMailmessage

Concept listUser IDs

View datastructure

(see Figure 2b)Learning feedback

6. Read e-mail; follow OM links

HTML-aware e-mail recipient client

2. View generationkMail server

(Existing) KnowledgeManagement System

Figure 2. kMail (a) process and (b) the view the data structure returned to the kMail sender client.

Table 1. Memory type categories.

CODE MEANING

1 Definition2 Graphic image (picture, schematic)3 Policy statement or guideline4 Specification (product, equipment)5 Opinion

[[Concept_1, [(Desc_1, Type, View, URL_1) , …, (Desc_n, Type, View, URL_n)],Concept_2, [(Desc_1, Type, View, URL_1) , …, (Desc_n, Type, View, URL_n)],…Concept_n, [(Desc_1, Type, View, URL_1) , …, (Desc_n, Type, View, URL_n)]]

(b)

(a)

Page 5: Tying Knowledge to Action with kMail - BIUteenid/publications/kmail.pdfThe kMail par adigm g rew from a resear ch project we undertook to investigate ways to improve intercultural

• Role(PersonID, StartDate, EndDate,Role),

• Supervisor(PersonID, StartDate, End-Date, SupervisorID),

• Project(PersonID, StartDate, EndDate,Project),

• Concept(ConceptID, ConceptName,ConceptDescription), and

• ConceptUsage(ConceptID, MemoryID,PersonID, Instance).

Four views of organizationalknowledge

Given the OM views’ basic structure, andthe way the system uses them,kMail canapply multiple views in determining whichmemory items the concept in the current e-mail message should associate with. Four ofthese views are personal,supervisory, proj-ect-related, and role-related.

View 1—personal. A person will prefer touse memory–concept associations that he orshe has used in the past. To determine thepersonal MCA view, we query the OM meta-knowledge to determine if the sender, asidentified by his unique PersonID (derivedfrom the sender’s e-mail address),has one ormore memory items that he has explicitl yassociated with a concept.

For a given concept C, a person mighthave zero, one, or more memory itemsassociated with that concept. This view pre-sents a memory items list that the user has,at one time or another, associated with thegiven concept. Clearly, this cannot be a suf-ficiently tight restriction on the searchspace. It is,however, a viable starting pointfor further refinement of the result set. Thishypothesis follows Eleanor Rorsch’s the-ory of typicality in that it assumes a con-sistency of usage across situations for thesame individual.10

View 2—supervisory. A memory–conceptassociation of a person’s immediate super-visor is preferable to a colleague’s mem-ory–concept association on the same projectand less preferable to a person’s own per-sonal MCA.

For a given concept C, the supervisor canhave zero,one, or more memory items asso-ciated with that concept. This view shows alist of memory items that the supervisor has,at one time or another, associated with thegiven concept.

Views 3 and 4—project-related and role-related. We can use a colleague’s MCA witha different role but on the same project toeffectively restore context. Alternatively, inView 4, a colleague’s MCA with the samerole on a different project,while less prefer-able than a same-project MCA,can be use-ful as well.

We use these first four views as a guide indetermining the sort order of memory itemresults returned to the system for linking withe-mail concepts. After we execute each of thefour queries,we merge and sort the resultsby MemoryType, ranking them within eachmemory item type according to the corre-sponding MCA. The system defaults to applythe highest rank to definition-type memoryitems,and continues down the list shown inTable 1. The actual order of presentation is acontrollable parameter, considering that dif-ferent users prefer alternative ranking orders.The latter three views are modeled after thetheory of conceptual coherence.11 We canalso create additional views,such as depart-mental,in a similar manner, as described inthe “Enhancing an e-mail message” sidebar.

Where do we go from here?

With kMail, the act of composing an e-mail message becomes integrated with theknowledge-management process unobtru-sively. Whenever a manager links a knowl-edge item to a message concept,he or she isexplicitl y confirming that item’s currentvalidity. By tracking the usage (and non-usage) of memory items,we can effectivelydetermine which OM parts are relevant anduseful.

The editing controls (Figure B in the side-bar) give an e-mail author an intuitive inter-face for knowledge management. At anypoint in validating the memory items that thesystem selects for inclusion in the e-mail,thesender can create a new OM entry and mod-ify or delete any of the entries that the sys-tem can retrieve. Modifications and dele-tions,however, occur only at the view level,so the underlying OM remains intact and weonly update the metaknowledge to reflect theuser’s input. The exception to this immedi-ate update is when the user adds a new mem-ory item to the system. An entry of a new def-inition or a new URL inclusion might berelevant to the entire organization. A mech-anism to manage this sort of knowledgemaintenance is the subject of future research.

Without an organizational-validation mech-anism,the author in question must keep thenew additions until such time as he or she canapprove them for broader access.

Learning. Whenever the user sends an e-mail message with an MCA link that he orshe has explicitly verif ied, this informationgoes to the server to update the user profile.We consider explicit verif ication the simpleact of viewing an MCA link to a given con-cept and leaving it in place. This “non-act”provides confirmation that the user acceptsthe default association that the systemchooses. If the sender explicitly changes theMCA link, replacing it with an alternative,this information is transmitted as well. Notonly can this information update the userview, it also provides important feedback asto whether the chosen OM views and theorder in which they are applied are appro-priate. By monitoring the ways in whichknowledge links up with action in practice,kMail takes a step toward being a participa-tory8 KMS, alleviating the danger of provid-ing outdated materials.

Limita tions. As in any KMS component,there are two nonsystemic functions that arevital to the system’s success yet difficult tocontrol. The first,common to all knowledge-management environments,is a sufficientlevel of user participation to update knowl-edge and keep it current. kMail takes a steptoward promoting this participation by pro-viding update functions to the e-mail author.But,when a manager is composing an action-oriented e-mail message, he wants that mes-sage to be clarif ied and sent—and that is notthe most opportune time to be updatingknowledge. In this respect,kMail suffers fromthe same shortcoming as any other KMS.

A second, kMail-specific limitation relatesto appropriate views creation and use. As wehave shown, the system’s successful usedepends on properly defining the views rele-vant to each user. This, too, is in the user’shands,and although we have based initialviews on viable contextualization theories,there is clearly more work we could dohere.10,11

Futur e directions. Updating the underlyingOM is possible by trickling down modifica-tions from the metaknowledge whenever auser has initiated a change. This requiresestablishing further validation mechanisms,because what the user is changing is his orga-

MAY/JUNE 2000 37

Page 6: Tying Knowledge to Action with kMail - BIUteenid/publications/kmail.pdfThe kMail par adigm g rew from a resear ch project we undertook to investigate ways to improve intercultural

nizational knowledge view, not necessarilythe underlying memory items in the OM.

Incorporating recipient feedback is anotherdirection in which we could extend this work.

CONNECTING KNOWLEDGE TO AC-tion through e-mail provides a two-way streetfor knowledge management on the Internet.It allows relevant OMs delivery directly toan action point,while it also creates a naturalsetting in which users can update and main-tain the system as part of their daily e-mailinteractions.

The use of memory–concept association,a metaknowledge-based associative memory,provides a series of useful,extensible asso-ciations between different users and the OMknowledge. Managing the MCAs through a

straightforward view mechanism lets usallow external access to an organization’sinternal memory items with sufficiently finelevels of control.

The kMail development,with its roots indealing with communication problems inmultinational corporations,opens new, prac-tical directions for knowledge managementand dissemination across the Internet.

AcknowledgmentsWe thank Ehud Dayan for developing the kMail

code and Delta Textiles Ltd. for cooperating withour field study. This work was supported by a grantfrom the Chase Manhattan Bank.

References1. D.G. Schwartz,“When E-Mail Meets Organi-

zational Memories,” Int’ l J. Human-ComputerStudies,Vol. 51,No. 3,1999,pp. 599–614.

2. D. Te’eni and D.G. Schwartz, “Contextual-ization in Computer-Mediated Communica-tion,” Information Systems—The Next Gen-eration, L. Brooks and C. Kimble, eds.,McGraw-Hill, New York, 1999,pp. 327–338.

3. J. Habermas,The Theory of CommunicativeAction, Beacon,Boston,1981.

4. D.E. O’Leary, “Using AI in Knowledge Man-agement:Knowledge Bases and Ontologies,”IEEE Intelligent Systems, Vol. 13, No. 3,May/June 1998,pp. 34–39.

5. C.A. O’Reilly and L.R. Pondy, “Organiza-tional Communication,” Organizational Behav-ior, S. Kerr, ed.,Grid, Columbus,Ohio,1979,pp. 119–150.

38 IEEE INTELLIGENT SYSTEMS

Enhancing an e-mail messageFigure A shows a portion of the original e-mail message com-

posed by a product line manager regarding the plans for the upcom-ing line of men’s trunks.

The first processing stage produces the same letter with

embedded links and related potential links. Figure B showshow a departmental-analysis view affects the “exposed elastictab” knowledge. In this instance, the view presents memoryitems based on department of origin (Production,Design,Sales,Finance). The initial link chosen is the fabric sample,indicated in red, as the preferred organizational memory of thee-mail author. Here the system presents the results grouped bydepartment,following the parameters that the sender sets. TheFigure A. Original e-mail message.

Figure B. Default personal view, exposed elastic tab trunks.

Original email message Edit and Current default memorywith embedded links validation controls item in view

OM memory-item display Results of multiple OM viewsfor given concept .

Page 7: Tying Knowledge to Action with kMail - BIUteenid/publications/kmail.pdfThe kMail par adigm g rew from a resear ch project we undertook to investigate ways to improve intercultural

6. A. Abecker et al.,“Toward a Technology forOrganizational Memories,” IEEE IntelligentSystems, Vol. 13,No. 3,1998,pp. 40–48.

7. M.S. Ackerman,“Definitional and Contex-tual Issues in Organizational and GroupMemories,” Proc. 27th Hawaii Int’ l Conf.System Sciences, IEEE Computer Soc. Press,Los Alamitos,Calif., 1994,pp. 203–224.

8. S. Buckingham Shum,“Negotiating the Con-struction and Reconstruction of Organiza-tional Memories,” J. Universal Computer Sci-ence, Vol. 3,No. 8,1997,pp. 899–928.

9. C.J. Date and H. Darwen, A Guide to the SQLStandard, Addison-Wesley, Reading, Mass.,1993.

10. E. Rorsch, “Principles of Categorization,”Cognition and Categorization, E. Rorsch andB.B. Lloyd, eds.,Lawrence Erlbaum Assoc.,Hillsdale, N.J., 1978.

11. G.L. Murphy and D.L. Medin,“The Role of

Theories in Conceptual Coherence,” Psycho-logical Rev., Vol. 92, No. 3, 1985, pp.289–316.

David G. Schwartz is an assistant professor at theGraduate School of Business Administration ofBar-Ilan University, Israel,where he serves as headof the Information Systems Division. His researchinterests include Internet business applications,knowledge management,distributed AI, and theuse of management models for intelligent agentarchitectures. He serves as editor of the Journal ofInternet Research and is co-editor of Internet-Based Knowledge Management and Organiza-tional Memories. He is also a founding partner ofApropos IT Ventures,a venture capital fund focus-ing on early-stage information technology com-

panies in Israel. Contact him at the Center forGlobal Knowledge Management,Graduate Schoolof Business Admin., Bar-Ilan Univ., Ramat-Gan,52900,Israel; [email protected] or [email protected].

Dov Te’eni is the director of the Graduate Schoolof Business Administration at Bar-Ilan Universityand professor of information systems. His researchinterests include human–computer interaction fordecision making, decision-support systems,behav-ioral aspects of information systems,and softwareerrors. Contact him at the Center for GlobalKnowledge Management,Graduate School ofBusiness Admin., Bar-Ilan Univ., Ramat-Gan,52900,Israel; [email protected].

MAY/JUNE 2000 39

e-mail author can choose to link all,one, or a view combinationresults to the concept in the message. Figure Cwith production-related memory items and Figure D with design-related memoryitems show alternative OMs that the e-mail authorcan choose toinclude in the view to be sent.

Clearly, there are innumerable figuratively, not computation-ally ways to combine the user profile characteristics in determin-ing relevancy. For example, a query for same role, same projectmight be an effective combination in situations where projectshave multiple people filling the same role. There can be severalalternate views to the departmental focus shown in our example.

The OM views can be a powerful tool in properly utilizing theOM. Domain-specific criteria help determine what the relevant

views are in a given organization. Project-oriented companies willrequire a same projectview; an organization aligned using func-tional management approach will require functional views; and soforth.

kMail is Internet-friendly in its use and resources allocation.By treating all knowledge resources as URL-accessible, we elimi-nate any need for replication and redundancy (aside from the useof common proxy caching to improve performance). Knowledgecan be distributed strictly on an as-needed basis,and if we get par-ticular about it,on a need-to-know basis as well. Need-to-knowknowledge distribution is particularly relevant in a security-con-scious or competitive environment.

Figure C. Production view, exposed elastic tab components.

Figure D. Design view, exposed elastic tab original sketch.