chapter 12: synthesis & exam preparation · key ideas in behaviorism ① psychology is becoming...
TRANSCRIPT
CS-411 : Digital Education & Learning Analytics
ÉC OLE POL Y TEC H NIQ U EFÉ DÉRA LE D E LA USANNE
Pierre Dillenbourg, Patrick Jermann & Łukasz Kidziński
Chapter 12:
Synthesis & Exam preparation
Behavorism
Constructivism
Socio-cultural theory
MOOCs
DataViz
Learning Analytics
CS-411 : Digital Education & Learning Analytics
Orchestration Graphs
Behavorism
Constructivism
Socio-cultural theory
MOOCs
DataViz
Learning Analytics
CS-411 : Digital Education & Learning Analytics
Orchestration Graphs
Burrhus Frederic Skinner (1904-‐1990), Operant Conditonning
h>ps://sites.google.com/a/adams12.org/harp-‐mrhs-‐home-‐v1/06-‐powerpoint-‐secHon/skinner-‐s-‐operant-‐condiHoning
STIMULUS
STIMULUS
Behaviour
Association
How do people learn ? By condiHonning
p (B | S) analytics
Key ideas in behaviorism
① Psychology is becoming more scientific
② The brain is a black box; the focus is on behaviors
③ Learning is « engineered »
④ Association results from immediate feedback
⑤ The learner is permanently active
⑥ Small steps increase the probability of positive feedback è Programmed instruction
⑦ Then it moved to larger step: modular instruction è mastery learning
a1 a2..1 a2.2
Pre-requisites test
Remediation
a3
a4
a5
a6
a7
a8
Pre-test
Post-Test
Int. Test
Redo [a4-a5]
Skip [a4-a5]
Item Order
Diff
icul
ty
Uphill strategy: go up until he fails
Downhill strategy: go down until he succeeds
Discriminant strategy: increase/decrease difficulty based on success – cut the space in 2
Modular Instruction Pre-requisite test: Does the learner has the pre-requisite to start the course ? Pre-test: Should the learner skip some modules ? Intermediate-test: Didthe learner reach the objectives of this module ? Post-test: Did the learner reach the objectives of this course?
Adaptive Testing
h>p://arrowthroughthe
sun.blogspot.ch/2012/07/if-‐ph
ilosoph
y-‐ne
twork-‐is-‐based-‐on
.htm
l
Instructional Design
Knowledge space
Digital courses
Example of exam question
Which learning mechanism(s) benefit from immediate
feedback ? Why ?
① Elicitation
② Proceduralisation
③ Association
④ Metacognition
Example of exam question
For which learning objectives would you use a behaviorist
approach ? How could it work ?
① Identifying categories of clouds (cumulus, stratus,..)
② Identifying the bug in a piece of code
③ Identifying the best business plan among 3 proposals
④ Identifying a fossile on a stone
Behavorism
Constructivism
Socio-cultural theory
MOOCs
DataViz
Learning Analytics
CS-411 : Digital Education & Learning Analytics
Orchestration Graphs
experience behaviour
Constructivism
How do people learn ? By construcHng cogniHve structures from experience (trial & error)
Reflection
Trial & Error
h>p://fr.slideshare.net/susanhansen1460/une-‐psy250-‐session-‐7-‐ist-‐2-‐years-‐language-‐piaget
Dr Susan Hansse, University of New England
Cognitive Conflict as key learning mechanism
(1) I wanted to get this
(2) I got that
(3) The problem is here
define "house1 [[ ] [forward 100 right 45 forward 60 right 120 forward 60 right 45 forward 100 right 90 forward 60 ]]
METACOGNITION
Cognitive Conflict as key learning mechanism
• Learning from experience • Learning by doing • Learning from failure • Discovery learning
Conditions: 1. The conflict is detected 2. The learner finds how to solve it
Role of the environment (sequence of projets / teacher / peer)
Constructivism
Constructionism Guided Discovery
Microworlds
• microworlds • Simulations • Modelling
Radical Quest for effectiveness
Learning for simulations is difficult Hypothetico Deductive Reasoning
1. (Raise a question)
2. Generate an hypothesis
3. Design an experiment
4. Run/simulate the experiment
5. Interpret results
Manipulating real or virtual objects ?
Summary: From Constructivism to Augmented Reality
1. People don’t learn by being taught but by adapting their knowledge structures through interaction with artefacts. Educational philosophy: from telling students what to do to letting them invent things
2. In practice, this approach does not work very well without external support and requires talented teachers. Learning from simulation requires inquiry skills. Training these transversal skills are key goals of any education
3. Evolution of pedagogical methods from building mental schemes to building concrete objects. Digital artefacts offer rich interactions but digital education is not limited to virtual object. Tangible interfaces and augmented reality open it to physical
manipulation.
Example of exam question
How does assimilation and accomodation occur when
learning from a simulation ?
Example of exam question
For which learning objectives would you use a microworld ?
What would learners do in this microworld
① Learning to express clearly one’s idea
② Learning to decompose a problem
③ Learning to learn from one’s own mistakes
④ Learning to take the viewpoint of someone
Behavorism
Constructivism
Socio-cultural theory
MOOCs
DataViz
Learning Analytics
CS-411 : Digital Education & Learning Analytics
Orchestration Graphs
Social Interaction
Thinking
Inte
rnal
isat
ion
We internalise social interaction because thinking is a dialogue with oneself .
Private speech (Vygostky) Egocentric speech (Piaget)
Collaborative learning
1. Collaborative learning is often effective, but not systematically.
2. It is effective when rich interactions occur such as explanation, argumentation, mutual regulation
3. To make it more effective, the technology or the script increases the necessity for students to produce these interactions
4. The theory behind emphasizes that cognition is inherently social because thinking mostly relies on language.
Social Cognition
The hardware is individual
But the software is social
Example of exam question
There are 400 architectures students in an online course on urbanism. You would like to apply the JIGSAW script to design their team project. The project concerns the positioning of car parks in the city. What would the script do?
Example of exam question
For which learning objectives would you collaborative
learning methods ?
① Solving problems when there is a single clear solution
② Solving problems when there a several solutions, but
some are better than others.
③ Solving problems where the goal is to find many
solutions.
Behavorism
Constructivism
Socio-cultural theory
Learning Theory
Lesson (eLearning)
Exploratory environment
Collaboration Tools
Learning Technology
Behavorism
Constructivism
Socio-cultural theory
MOOCs
DataViz
Learning Analytics
CS-411 : Digital Education & Learning Analytics
Orchestration Graphs
400'000+%Users%%
195'000+%Users%%
50'000+%Users%%3'200+%Users%%
0%%
10%%
20%%
30%%
40%%
50%%
60%%
Watched%Video(s)% Solved%Exercise(s)% Passed%the%Course% Paid%$50%for%CerFficate%
%"of"R
egistered"Users"
55’000 X 4 ECTS ≈ 800 * 300 ECTS
MOOC BA+MA
The Gartner Cycle of Technology Adop8on
MOOC > VIDEOS
A good MOOC
includes rich activities
h>p://128.178.27.98:8082/LHE1.html
Fluid Dynamics (Gallaire & Ancey)
StaHcs (Mu>oni & Burdet)
Behavorism
Constructivism
Socio-cultural theory
MOOCs
DataViz
Learning Analytics
CS-411 : Digital Education & Learning Analytics
Orchestration Graphs
Données par canton Eléments visuels
• Canton • # inhabitants • % foreigners • % votes
• Label • Disk size • Position Y • Position X • Coulor >50%
A visualisation grammar is a mapping between information pieces and visual elements
3D trap !
0"
10"
20"
30"
40"
50"
60"
Janvier" Février" Mars"" Avril"" May"" Juin" Juillet"
Geneve%
Geneve"
0"
20"
40"
60"
Janvier"Février"
Mars""Avril""
May""Juin"
Juillet"
Geneve%
Design Principles
#3. Check graphical integrity (Tufte)
visual ~ data?
0
10
20
30
40
50
60
1 2 3 4 5
0
10
20
30
40
50
60
1 2 3 4 50
10
20
30
40
50
60
1 2 3 4 5
30
22
45
12
55
0
10
20
30
40
50
60
1 2 3 4 5
A B
C D
Design Principles #3. Optimize the“data ink ratio” (Tufte)
Which % of pixels provide information ?
h>p://med
ia.ju
iceanalyHcs.com/im
ages/smallm
ulHp
les1.png
Design Principles
#6
Use“small multiple”
(Tufte)
Design Error 4
Let R generate the scale automatically !!!!!!!!!
2014 Geneve Vaud Valais0 Fribourg Neuchatel 2015 Geneve Vaud Valais0 Fribourg NeuchatelJanvier 100 100 100 100 100 Janvier 50 50 50 50 50Février 80 39 140 110 120 Février 40 19.5 70 55 60Mars0 100 50 120 90 120 Mars0 50 25 60 45 60Avril0 100 100 100 100 100 Avril0 50 50 50 50 50May0 100 120 100 100 100 May0 50 60 50 50 50Juin 70 145 100 100 90 Juin 35 72.5 50 50 45Juillet 50 20 120 110 150 Juillet 25 10 60 55 75
600 574 780 710 780 3444 300 287 390 355 390 1722
2014 2015
00
200
400
600
800
1000
1200
1400
1600
Janvier0 Février0 Mars00 Avril00 May00 Juin0 Juillet0
Geneve0
Vaud0
Valais00
Fribourg0
Neuchatel0
00
100
200
300
400
500
600
700
800
Janvier0 Février0 Mars00 Avril00 May00 Juin0 Juillet0
Geneve0
Vaud0
Valais00
Fribourg0
Neuchatel0
Humanities
Management
Basic SciencesLife sciences
Engineering
HumanitiesManagementBasic SciencesLife sciencesEngineering
Design Erreur 10
‘split attention effect’
Vertical / Horizontal: 5:1
Vertical / Horizontal: 1:1
Example of exam question
Is there a split a>enHon effect on this figure ? Is there any other concern ?
Behavorism
Constructivism
Socio-cultural theory
MOOCs
DataViz
Learning Analytics
CS-411 : Digital Education & Learning Analytics
Orchestration Graphs
Independent Variable Variable
Dependent
Solo / Team Test Score
How to prove the effectiveness of X ?
Summary • Research Question To be answered by the experiment • Hypothesis Expected results ( A > B); an affirmation • Independent variables What one varies between the conditions (or Factor) • Modality Value of a factor • Condition Set of (factor, modality) per group of subjects • Control group The reference against which one will compare • Dimension Number of factors • Dependent variables How does one measure the effects ? • Controlled variables Things you try to keep constant or to randomize • Intermediate variables Explain the link from Independent to Dependent Variables • “Significant” difference Probably (<5%) not due to sampling error • Interaction effect The effect of one IV on the DV depends upon another IV • Between/Within subject Do subjects pass in one or several conditions ? • Counterbalancing Inverting the order of conditions for within-subject plans
Example of exam question
Design an experiment that compares the effectiveness of constructionism versus constructivism. Verify if the effect depends upon the age of students. What would be the independent, controlled, intermediate and dependent variables? Describe a potential interaction effect. Design an experiment that measures the effectiveness of immediate feedback from learning vocabulary. Verify if the effect varies with the level of prior knowledge of the participants. What would be the independent, controlled, intermediate and dependent variables ? Describe a potential interaction effect.
Behavorism
Constructivism
Socio-cultural theory
MOOCs
DataViz
Learning Analytics
CS-411 : Digital Education & Learning Analytics
Orchestration Graphs
Pairs of students with conflicting viewpoints, probably get higher learning gains (than …), because conflict increases argumentation intensity and opportunities for decentration.
Learning Theory (behaviorism, constructivis, socio-cultural)
prediction explanation
How to form conflicting pairs among 20, 200 and 2’000 students ?
Orchestration Graphs
G= (V, E) where E= V X V
V = {ai} | ai: ts, te, π, object, product, {c}, traces, {metadata}
E = { eij} | eij: (ai, aj, {operators}, {controls}, label, weight, elasticity)
An orchestration graph is a weighted directed geometric graph.
Library of Edge Labels Why is ai a condition for aj ?
The preparation edges connect two activities when the learner has a higher probability of succeeding at aj if he carried out ai before ai.
The set edges connect two activities when the skills or contents addressed in ai and aj are in relationship with each other; for example, subset/superset, whole/part, and siblings. (UP / DOWN)
The translation edges connect two activities in which the same content is addressed under different formats, representations, notations, or viewpoints. Learners therefore have to translate the representation used in ai into the representation used in aj.
The generalization edges introduce variations of the content or skills across the space of generalization, namely introducing the student to more general, less general, or analogical contexts from ai to aj. (UP / DOWN)
Library of Edge Labels Why is ai a condition for aj ?
Preparation Set Translation Generalization
(P) Prerequisite (S+) Aggregation (T) Proceduralization (G+) Induction
(P) ZPD (S+) Expansion (T) Elicitation (G+) Deduction
(P) Adv. organizer (S–) Decomposition (T) Alternate (G+) Extraction
(P) Motivation (S–) Selection (T) Reframe (G+) Synthesis
(P) Anticipation (S=) Juxtaposition (T) Reverse (G=) Analogy
(P) Logistics (S=) Contrast (T) Repair (G=) Transfer
(P) Data collection (S=) Identity (T) Teach (G–) Restriction
Library of Edge Operators
How data collected in ai are processed for aj ?
Aggregation operators gather data for subsequent activities, generally located on a higher plane
Distribution operators split data for subsequent activities, generally located on a lower plane
Social operators modify the social structure of activities. They rely on social distance criteria
Back-office operators enrich data with external information, including information manually provided by human actors
Library of Graph Operators
Aggregation Distribution Social BackOffice (A) Listing (D) Broadcasting (S) Group
formation (B) Grading
(A) Classifying (D) User selection (S) Class Split (B) Feedback (A) Sorting (D) Sampling (S) Role assignment (B) Anti-plagiarism (A) Synthesizing (D) Splitting (S) Role rotation (B) Rendering (A) Visualizing (D) Conflicting (S) Group rotation (B) Translating (D) Adapting (S) Drop out
management (B) Summarizing
(S) Anonymisation (B) Converting (B) Updating
Learner Modelling
Text entered Item selected (button, menu,…) Area clicked Line drawn with mouse. Response time … … Number of pauses Mouse path Gaze path Facial expressions Gestures …
From the learner’s behaviours, infer his/her learner’s knowledge state
Behaviours
Behavioural ‘Dust’ (fragments of behaviour that do not
have an explicit semantic value)
« Social Signal Processing »
a1
a2
a3
X0(s) X1(s) X3(s)
X2(s)
Dropped Out
Active Lost
.05 Fine
.17 .27
.40
.39 .39
.10
.24
.14
.30
.35 .20
1
The weight of edges
Plane of Activity π1 π2 π3
Plan
e of
Mod
elin
g
Indi
vidu
al M
odel
(π
1)
Xi (s) Xi (s1) Xi (s1) Active / Passive Social loafing With-me
On-Leave / Drop-
Out / Late-Comer
Free-rider Central
Disoriented Leader Isolated
Linear rigidity On/Off Role Bridge
Impasse
Trapped
Over/Under
generalization
Deep/surface
Gaming
Gro
up M
odel
(π
2)
Xi (s1,s2,s3,… ) Xi (s1,s2,s3,… ) Under/Over Sized Cluster
Cognitive/Emotional
Conflict
Misunderstanding
Groupthink
Distributed
Clas
s M
odel
(π
3)
Xi (S) Good/Bad Spirit
Slow
Split
Libr
ary
of S
tate
s
Example of exam question
A pedagogical scenario starts with a test on english vocabulrary. Then, the high score students have a lecture on the construction of negative sentences, while the low score do indivdiual exercies to acquire more vocabulary. Then, students work in pair, one low score and one high score on negative sentence construction exercices Describe the graphs and specify the label and oeprators on edges.
Example of exam question
Invent a graph with an edge ‘alternate’ and 2 operators, aggergation and distribution.
xi(sb)
bi(sa)
xi(sa)
bi(sa) bi-1(sa)
Time
Soci
al
Diagno
sis xi-1(sb)
xi-1(sa)
Δx i
-1 Δ
xi
The diagnosis cube