At the Social Simulation and Serious Games special track at ESSA 2014, Setsuya Kurahashi gave this talk on the effects of different kinds of collaborative learning on different kinds of school classes.

1. 1
Teaching Simulation on Collaborative Learning,
Ability Groups and Mixed-ability Groups
Setsuya KURAHASHI* and Keisuke KUNIYOSHI
Graduate School of Systems Management (GSSM)
University of Tsukuba, Tokyo

2. 2
Agenda
• Motivation and Aims
• Related Work of Learning theory
• Item Response Theory
• Graphical Test Theory
• Complex Doubly Structured Network
• Learning Model with Complex Doubly Structured Network
• Experiment 1 : Effect of teaching strategies
• Experiment 2 : Effect of collaborative learning
• Experiment 3 : Effect of seating arrangement
• Experiment 4 : Effect of ability classes
• Discussion and Summary
• Future work

3. 3
Development of Human Resources
Cultural
Capital Education
Development
of Human
Resources
Gene
MOOC

4. Motivation and Aims
• What kind of influence could teaching
strategies have on learning effects?
• Modeling of a learning process of each student and
teaching strategies.
• What kind of influence could the seating
arrangement of learners have on
collaborative learning effects?
• Modeling of learner’s interaction in a classroom.
• What kind of influence could ability groups
and mixed-ability groups have on
collaborative learning effects?
• Scenario analysis of learning environments.
• The aims of the model is to analyse the actual
conditions of understanding of learners
regarding instructions given in classrooms.
4

5. Related Work of Learning Theory
Test
theory
This study
Item Response
Theory (IRT)
5
Graphical Test Theory
Bayesian Network
Learning
structuration
study
• Learning Material
Structure Analysis
• Course Outline
Determination
• Item Relational
Structure (IRS)
Complex Doubly
Structural Network
Social
network
study
Doubly Structural
Learning Model
Probabilistic
reasoning
method

6. The approach of this study
6
Test theory for exam questions (IRT)
Learning material structure model (Bayesian net)
Collaborative learning approach
In-class learning process regarding a teaching
strategy is one of unexplored fields.
Quantitative method of collaborative
learning has not been developed yet.
■This study
The
understanding
status,
knowledge
structure,
and
collaborative
effect
of
each
learner
are
simulated
on
an
agent-based
model
integrated
by
using
a
complex
doubly
structural
network.
Experiment 1: Effect of teaching strategies in a classroom
Experiment 2: Effect of collaborative learning
Experiment 3: Effect of seating arrangements
Experiment 4: Effect of ability classes

7. Probability of Understanding
Item Response Theory : IRT
7
Item
Response
Theory
(IRT)
has
been
proposed
to
evaluate
examina9on
ques9ons.
• IRT
• Standard Test Theory
• n Parameters Logistic Model
• Xn : exam question n
Estimation of Item Parameters(2PL)
2PL Model
θ =
+
−θ −

8. 8
Graphical Test Theory: Bayesian Network
Learning
item
Ｘ１
Ｘ2
Ｘ３
Ｘ4
Ｘ５
■Estimation of a learning material structure
fx3 = P(X1)×
P(X1, X3)
P(X3)
= 0.82
fx1 = P(X1)
fx2 = P(X1)× P(X3)× P(X4)×
P(X2, X1, X3, X4)
P(X1, X3, X4)
= 0.74
fx 4 = P(X1)× P(X3)
×
P(X4, X1, X3)
P(X1, X3)
fx5 = P(X2)× P(X3)× P(X4)
×
= 0.83
P(X5, X2, X3, X4)
P(X2, X3, X4)
= 0.61
Dependency
relationship
Conditionally probability of dependency

9. 9
Complex Doubly Structural Network
Internal Network
(knowledge)
Social Network
(society)
Internal Network
This model can express networks in microscopic
and macroscopic ways as an integrated model.

10. Doubly Structural Learning Model
10
Understanding
probability
Teacher
Social Network
Internal Network
Social (Classroom) Network
Students
Teaching
Level of achievement
Knowledge
structure
Classroom
Internal (Knowledge)
NetworkTeaching strategy
Knowledge
Collaborative
learning
2
)
2
*
2
+
2
,
2
-
) *
Teacher
+ , -
. / 0 1 )(
)) )* )+ ), )-
). )/ )0 )1 *(
*) ** *+ *, *-
*. */ *0 *1 +(
'

11. !
#
$
Item Response Theory
%

12. Simulation Method
11
■In-class learning model
・30 learners in a classroom
・5 teaching materials: X1, X2, X3, X4, X5
■Internal network
From arithmetic exam answer data of 300
learners, estimating (1) understanding
probability(IRT), (2) material structured
model(Bayesian Network)
■Social network
From seating allocation and correct
answer data in a class, modeling a
social network in a classroom.
■Simulation
This simulation is to estimate what material should be taught, in what order and
how many times, until all learners in the classroom could give the correct
answer.
■Criteria
1) Attainment degree : the proportion of correct answer
2) Average teaching time : the time until the attainment degree has reached 1

13. Estimated Ability and Understanding
Probability
12
Item parameters: ability, difficulty, discrimination (IRT)
Understanding probability (IRT, Bayesian network)
Correct or Incorrect
Ability
Probability
X1
X2
X3
X4
X5
X1
X2
X3
X4
X5
1
1
1
1
1
1
0.8457
1
0.93
0.96
0.95
0.68
2
1
1
1
1
0
0.1658
0.99
0.77
0.89
0.87
0.39
3
1
1
1
0
1
0.2427
0.99
0.8
0.9
0.88
0.43
4
1
1
1
0
0
-0.297
0.97
0.57
0.79
0.76
0.22
5
1
1
0
1
1
0.2
0.99
0.78
0.89
0.87
0.41
6
1
1
0
1
0
-0.332
0.97
0.56
0.78
0.75
0.21
7
1
1
0
0
1
-0.268
0.98
0.59
0.8
0.77
0.23
8
1
1
0
0
0
-0.731
0.91
0.37
0.64
0.63
0.12
9
1
0
1
1
1
0.0953
0.99
0.74
0.88
0.85
0.37
10
1
0
1
1
0
-0.419
0.96
0.52
0.75
0.73
0.19
…
…
…
…
…
…
…
…
…
…
…
…
29
0
0
0
1
1
-1.102
0.78
0.22
0.49
0.49
0.07
30
0
0
0
1
0
-1.475
0.55
0.12
0.35
0.35
0.04
31
0
0
0
0
1
-1.427
0.58
0.13
0.36
0.37
0.04
32
0
0
0
0
0
-1.817
0.31
0.06
0.23
0.24
0.02

14. Simulation procedures
13
For each knowledge Xn {
For each learner {
If a learner's Xn == 0 {
if a collaborative relationship == true on the social network and a
student on the relationship already understand the knowledge {
P(Xn) - 1
} else {
Select subsequent knowledge connected Xn on the internal network.
The probability is calculated as multiplication of the probability of
all precedent knowledge connected and the conditional probability.
}
Based on the probability calculated, the value is conversed into two
values 1 and 0.
}
}
A=er
evalua9ng
understanding
status
of
all
students
for
every
material,
a
teacher
decides
the
most
effec9ve
material
to
teach
next.

15. 14
Transition of the understanding status

16. ODD protocol
15
Overview
1.Purpose
What kind of influence could teaching strategies have on learning effects?
What kind of influence could ability groups and mixed-ability groups have
on collaborative learning effects?
2.Entities, State valuables and Scales
Teacher: teaching strategy, understanding status of students
Students: understanding status,
3.Process overview and scheduling
Applying teaching strategies for some class patterns in which students are
seated, and then by comparing the average time of teaching sessions and
the attainment degrees.

17. ODD protocol
16
Design
concepts
4.Design concepts
・Basic Principles:
Item Response Theory for understanding probability model
Bayesian Network for the course material structure model
Social Network and the complex doubly structured network model
・Emergence: Collaborative learning
・Adaptation:
A teacher decides order of a teaching material depending on understanding
status of students.
・Objectives: Understanding status of students, teaching time
・Learning: No
・Prediction: A teacher predicts the most effective order of a teaching material.
・Sensing: Understanding status
・Interaction: Collaborative learning between students
・Stochasticity: Seating arrangements,
・Collectives: In-class social network between students and a teacher
・Observation: Understanding status, teaching times

18. ODD protocol
17
Details
5.Initialization
Teacher: 1, Student: 30, Material: 5
Teaching strategy: 4 types
Learning style: lecture, left-and-right, group
Seating arrangement: random, concentrated, dispersed
Ability class: mixed-ability, high ability, medium ability, low ability
6.Input Data
Arithmetic examination results of 300 students from an online learning system
7.Submodels
Item response theory
Graphical Test theory
Course material structure model
Complex doubly network theory

19. EXPERIMENT 1:
EFFECT OF TEACHING
STRATEGIES
18

20. Comparison between teaching strategies
TS No.
Method
TS 1 Teaching along with estimation by the
complex doubly structured network method
TS 2 Teaching by selecting items to teach in a
random manner
TS 3 Teaching an item where many learners gave
wrong answers
TS 4 Teaching by moving to next item when all
learners understood an item by order of the
highest correct answer rate
19
What kind of influence teaching strategies could have on learning effects.
■Teaching strategy
Teacher

21. 20
The transition of attainment degree
TS 1 TS 2
TS 3 TS 4

22. 21
Results of Experiment 1:
Teaching Strategies
■Simulation result （The average teaching time）
No.
Method
Lecture style
(Non-collaborative)
TS 1 Teaching along with estimation by the complex doubly
structured network method
22.5
TS 2 Teaching by selecting items to teach in a random
manner
41.4
TS 3 Teaching an item where many learners gave wrong
answers
32.3
TS 4 Teaching by moving to next item when all learners
understood an item by order of the highest correct
answer rate
23.4
TS 1 has the highest attainment degree, TS 4 is the second best which
adopts the teaching order of the highest correct answer rate.

23. EXPERIMENT 2:
EFFECT OF COLLABORATIVE
LEARNING
22

24. Collaborative Learning Model
23
What kind of influence collaborative learning could have on learning effects.
Left-and-right
collaborative
learning
Group
collaborative
learning
Teacher
Lecture style learning
(non-collaboration)
Teacher
Teacher
High academic
capability

25. 24
Results of Experiment 2:
Collaborative learning
■Simulation result （The average teaching time）
No.
Method
Collaboration type
1) Collaborative learning effect is higher than non-collaborative one.
2) Group collaborative learning is higher than left-and-right collaborative
learning.
Lecture
Left-and-right Group
TS
1
Teaching along with estimation by the
complex doubly structured network method
22.5 8.2 6.0
TS
2
Teaching by selecting items to teach in a
random manner
41.4 17.7 13.6
TS
3
Teaching an item where many learners gave
wrong answers
32.3 11.8 8.3
TS
4
Teaching by moving to next item when all
learners understood an item by order of the
highest correct answer rate
23.4 9.3 6.0

26. 25
EXPERIMENT 3:
EFFECT OF SEATING ARRANGEMENTS
IN COLLABORATIVE LEARNING

27. 26
Comparison between seating arrangements
What kind of influence the seating arrangement could have on learning
effects.
Concentrated
arrangement
Dispersed
arrangement
Teacher
High academic
capability

28. 27
Effect of seating arrangements
in “left-and-right” collaborative learning
Random arrangement
Concentrated arrangement Dispersed arrangement
Teacher
Teacher
Teacher

29. High academic
capability
Average teaching time: Left-and-right collaboration
Concentrated
Random
Dispersed
9.5
8.2
7.7

30. Effect of seating arrangements
in “group” collaborative learning
28
Random arrangement
Concentrated arrangement Dispersed arrangement
Teacher
Teacher
Teacher
Average teaching time Group collaboration
Concentrated
Random
Dispersed
8.4
6.0
5.6

31. Results of Experiment 3:
Seating arrangement
29
Average teaching time
Seating arrangement
Collaborative
learning type
Random Concentrated Dispersed
Left-and-right 8.2 9.5 7.7
Group 6.0 8.4 5.6
1) While the teaching time increases in the concentrated arrangement. it
decreases in the dispersed arrangement.
2) Learning effects vary by making changes in the seating arrangement
and the dispersed arrangement could enhance teaching effects.

32. EXPERIMENT 4:
EVALUATION OF EFFECTS ON
ABILITY CLASSES
30

33. 31
Evaluation of the effects on ability classes
What kind of influence ability classes could have on learning effects.
Lecture style
High ability class Medium ability class Low ability class
Average teaching time of ability classes
High
Medium
Low
17.0
20.0
23.7
Teacher
Teacher
Teacher

34. 32
Evaluation of the effects on ability classes
Left-and-right collaborative learning
High ability class Medium ability class Low ability class
Average teaching time of ability classes
High
Medium
Low
7.8
8.4
9.3
Teacher
Teacher
Teacher

35. 33
Evaluation of the effects on ability classes
High ability class Medium ability class Low ability class
Average teaching time Group collaboration
Teacher
Teacher
Group collaborative learning
High
Medium
Low
6.0
6.9
7.0

36. 34
Results of Experiment 4: Ability class
Total number of average teaching time for mixed-ability and ability classes
Learning type Mixed-ability
classes
Ability classes
Lecture
(non-collaborative)
67.5 60.7
Left-and-right
collaborative learning
23.1 25.5
Group
collaborative learning
16.8 19.9
In the lecture model, teaching time for the ability classes is less than the
mixed ability classes.
In the left-and- right, group collaborative learning model, teaching time for
the ability classes is more than the mixed ability classes.
The results indicate that ability classes have adverse effects on learners in
collaborative learning.

37. 35
Results of Experiment 4: Ability class
Average teaching time for mixed-ability and ability groups
Mixed-ability
class
Ability class
Learning style High Medium Low
Lecture 22.5 17.0 20.0 23.7
Left-and-right 8.2 7.8 8.4 9.3
Group 6.0 6.0 6.9 7.0
The ability classes for students of high academic capability are effective
more than or equal to mixed-ability classes, while not effective for students
of medium and low academic capability in collaborative learning.

38. 36
Discussion and Summary
• We designed the integrated simulation model for in-class learning
processes considering academic capability, leaning material structure and
collaborative relationship by interfacing internal and social network.
• 1st experiment: Effect of a teaching strategy
• Different teaching strategies cause different effects of learning.
• The proposed teaching strategy has the highest attainment degree,
• 2nd experiment: Effect of a collaborative learning
• Collaborative learning has a positive effect more than the lecture style.
• 3rd experiment: Effect of a seating arrangement
• A dispersed seating arrangement is more effectively than a concentrated seating
arrangement.
• 4th experiment: Effect of ability classes
• Mixed-ability classes are more effective than ability classes in the collaborative
learning, while ability classes are effective in the lecture style.

39. Discussion and Summary
37
How influence could teaching strategies have on learning effects?
(1) When different teaching strategies, seating arrangements, and
collaborative learning are used, learning effects vary,
(2) group style collaborative learning on dispersed seating
arrangements using the doubly structural learning model has high
learning effects, and the second best is the method in order of the
highest answer rate, and
(3)an ability group has negative effect on collaborative learning
because they reduce diversity in a class, so homogeneity between
learners has the risk to make collaborative effect fall into decline,
(4) whereas, if teaching is done one time for one knowledge item,
some learners could fall behind in the learning progress. Reviews
should be conducted repeatedly to facilitate the anchoring of the
knowledge in a class.

40. 38
Future work
• We were not concerned about negative effects from
unskilful students or misunderstanding, so we’ll add the
negative effects in the model.
• Some classes adopt more dynamic collaborative learning
where high ability students are allowed to walk around
and teach or discuss with others in a class. This dynamic
situation should be designed in our model.