MOTIVATIONALLY INTELLIGENT EDUCATIONAL SYSTEMS

1
MOTIVATIONALLY INTELLIGENT EDUCATIONAL SYSTEMS

• Benedict du Boulay
• Human Centred Technology Group
• University of Sussex
• UK
• Future e-Learning Conference 2008 - Istanbul

2
Thanks to

• Rose Luckin
• Teresa del Soldato
• Genaro Rebolledo Mendez
• Erika Martinez-Miron
• Amanda Harris
• and others in the HCT Group

3
Contents

• E-learning student motivation
• Motivationally intelligent systems
• An example
• Some questions and conclusions

4
e-Learning Three Aspects

• A style of learning with a particular focus on
  technology-mediated interactivity and
  collaboration
• The use of computer technology in leaning with a
  particular focus on internet technology
• The set of skills that enables learners to
  exploit technology in order to develop
  understanding or capability.
• http//www.reveel.sussex.ac.uk/

5
e-Learning Three Aspects

• A style of learning with a particular focus on
  technology-mediated interactivity and
  collaboration
• The use of computer technology in leaning with a
  particular focus on internet technology
• The set of skills that enables learners to
  exploit technology in order to develop
  understanding or capability.

6
My Perspective

• Artificial Intelligence in Education
• Intelligent educational systems
• Adaptation of system to the learner(s) via
  adjustments to e.g.
• Content, Communication, Help, Style
• Affective dimension of learning

7
The Goals of Expert Human Teachers

• . . . first, to sustain and enhance their
  students motivation and interest in learning,
  ... and second, to maintain their pupils
  feelings of self-esteem and self-efficacy, even
  in the face of difficult or impossible problems.
• (Lepper, Aspinwall, Mumme, Chabay, 1990)

8

• An intelligent educational system is a system
  that is able to deploy resources and tactics
  dynamically such that motivated students using
  the system will learn.

9

• An intelligent educational system is a system
  that is able to deploy resources and tactics
  dynamically such that motivated students using
  the system will learn.
• A Motivationally Intelligent educational system
  is an (intelligent) system that is able to deploy
  resources and tactics dynamically to maintain or
  increase the students desire to learn and her
  willingness to expend effort in so doing.

10

• An intelligent educational system is a system
  that is able to deploy resources and tactics
  dynamically such that motivated students using
  the system will learn.
• A Motivationally Intelligent educational system
  is an (intelligent) system that is able to deploy
  resources and tactics dynamically to maintain or
  increase the students desire to learn and her
  willingness to expend effort in so doing.

The two may suggest conflicting courses of action
11
(No Transcript)
12
Ways Of Thinking About Motivation

• What kinds of inner needs drive learning e.g.
  seeking approval and status
• What personal value is ascribed to learning
  process and outcomes
• What judgements are made about self-capability
• What stance is taken to the learning task
• How does the learner believe she is viewed by
  peers

13
Systems Need To Deal With

• curiosity, engagement, focus, exertion,
  confidence, control, flow, hope, pride,
  self-esteem, face
• frustration, disenchantment, boredom, perplexity,
  fatigue, disappointment, confusion

14
Cognitive and motivational model and reasoning
by system
Range of reactions by system

Range of sources of input from student pre- and
during session
15
Cognitive and motivational model and reasoning
by system
Range of reactions by system

Range of sources of input from student pre- and
during session
That nice feeling of understanding something
new That nasty feeling of being seen to not
understand
16
(No Transcript)
17
(No Transcript)
18
(No Transcript)
19
Ecolab2 reactions to challenge and help chosen
20
Motivational-Ecolab
21
Goal Orientated Ecolab
Feedback
Providing help
Elements of the interface
Performance Ecolab
Mastery Ecolab
22
(No Transcript)
23
INITIAL EVALUATIONS OF PRE/POST TEST
24
Research Questions

• How far can we trade-off between the categories?
• Can a model encompass all the categories?
• How important is the meta-level?
• Is a domain-free version feasible?
• Are there plausibility issues?

25
Conclusions

• We should take motivational issues seriously as
  part of future e-learning
• Motivationally intelligent systems have a wide
  variety of inputs and reactive tactics to deploy
• Choosing how best to operate in this space of
  possibilities is still a relatively open issue

26
Not Just Knowledge Skill

• curiosity, engagement, focus, exertion,
  confidence, control, flow, hope, pride,
  self-esteem, face
• frustration, disenchantment, boredom, perplexity,
  fatigue, disappointment, confusion

27
(No Transcript)
28
Motivational-Ecolab
29
(No Transcript)
30
Low persistence High confidence Failure
31
CONTENTS

• Theories of motivation
• Framework for motivationally intelligent
  educational systems
• Five research questions

32
Motivation

• The physiological process involved in the
  direction, vigour, and persistence of behaviour,
  Bergin, Ford Hess (1993)
• Willingness to expend time and energy on learning
• Quantitative - how far motivated?
• Qualitative - orientation, rationale
• Overlapping and multiple theories

33
Theories of Motivation

• Cognitive Evaluation Theory (Ryan Deci, 1990)
• Individuals deep-seated psychological needs and
  their learning activity e.g. approval
• Extrinsic vs intrinsic rewards
• Leverage tune rewards to learners perceptions
  and needs

34
Socio-Cognitive Theory

• Expectancy-value theory (Wigfield Eccles, 2000)
• Relation between choice of tasks and their
  persistence, vigour and performance
• Self-efficacy (Bandura, 1977)
• Self-judgement of capability to succeed
• Leverage work on learners self-understanding

35
Achievement Goal Theory

• Mastery Performance (Pintrich, 2000)
• Integration of motivation, affect and learning
• The emotions experienced during goal pursuit are
  sources of information that students interpret
  and use (Boekarts, 2003)
• Importance of social and contextual factors, e.g.
  adolescents (Wentzel, 1997)
• Leverage work on way learning activity is framed

36
Goal Orientated Ecolab
Feedback
Providing help
Elements of the interface
Performance Ecolab
Mastery Ecolab
37
(No Transcript)
38
(No Transcript)
39
(No Transcript)
40
(No Transcript)
41
MOTIVATIONALLY INTELLIGENT EDUCATIONAL SYSTEMS

• Benedict du Boulay
• IDEAs Lab
• Human Centred Technology Research Group
• University of Sussex
• UK

42
SOME LEARNER STEREOTYPES

• clever, confident but lazy
• clever, anxious, hard-working but lacking
  confidence
• performance oriented and hard-working
• weak and disengaged and does not make use of
  available help
• energetic but lacking focus tends to get
  distracted

43
CONTENTS

• Framework for motivationally intelligent tutoring
  systems
• Examples and evaluation of systems built at
  Sussex
• Conclusions

44
Thanks to the work of

• Rose Luckin
• Teresa del Soldato
• Genaro Rebolledo Mendez
• Erika Martinez-Miron
• and others in the HCT Group

45
Motivationally and affectively intelligent systems

• curiosity, engagement, focus, exertion,
  confidence, control, flow, hope, pride,
  self-esteem
• frustration, disenchantment, boredom, perplexity,
  fatigue, disappointment

46

• An intelligent system is a system that is able to
  deploy resources and tactics dynamically such
  that a motivated student using the system will
  learn.

47

• An intelligent system is a system that is able to
  deploy resources and tactics dynamically such
  that a motivated student using the system will
  learn.
• A Motivationally Intelligent system is an
  (intelligent) system that is able to deploy
  resources and tactics dynamically to maintain or
  increase the students desire to learn and her
  willingness to expend effort in so doing.

48

• An intelligent system is a system that is able to
  deploy resources and tactics dynamically such
  that a motivated student using the system will
  learn.
• A Motivationally Intelligent system is an
  (intelligent) system that is able to deploy
  resources and tactics dynamically to maintain or
  increase the students desire to learn and her
  willingness to expend effort in so doing.

May suggest conflicting courses of action
49

• An intelligent system is a system that is able to
  deploy resources and tactics dynamically such
  that a motivated student using the system will
  learn.
• A Motivationally Intelligent system is an
  (intelligent) system that is able to deploy
  resources and tactics dynamically to maintain or
  increase the students desire to learn and her
  willingness to expend effort in so doing.
• An Affectively Intelligent system is a
  motivationally intelligent system that provides
  an educational experience that leaves the student
  wishing to engage in further educational
  experiences.

50

• An intelligent system is a system that is able to
  deploy resources and tactics dynamically such
  that a motivated student using the system will
  learn.
• A Motivationally Intelligent system is an
  (intelligent) system that is able to deploy
  resources and tactics dynamically to maintain or
  increase the students desire to learn and her
  willingness to expend effort in so doing.
• An Affectively Intelligent system is a
  motivationally (intelligent) system that provides
  an educational experience that leaves the student
  wishing to engage in further educational
  experiences.

teacher
parent
May suggest conflicting courses of action
51

• Motivationally and affectively intelligent
  systems concerned with such issues as
• curiosity, engagement, focus, exertion,
  confidence, control, flow, hope, pride,
  self-esteem
• frustration, disenchantment, boredom, perplexity,
  fatigue, disappointment
• BUT
• Design time (e.g. nature of narrative)
• vs Runtime decisions (e.g. be polite or rather
  firm now)
• Cost benefit and tradeoffs of extra complexity vs
  educational effectiveness
• Long term traits and short term factors

52
Cognitive and affective model and reasoning by
system
Range of reactions by system

Range of sources of input from student pre- and
during session
53
Cognitive and affective model and reasoning by
system
Range of reactions by system

Range of sources of input from student pre- and
during session
That nice feeling of understanding something new
54
(No Transcript)
55
(No Transcript)
56
INPUT AND REACTION NOT NECESSARILY IN SAME
DIMENSION

• Sweaty palms get other students involved in the
  problem
• Lack of questions from student make task easier
  (or harder!)
• Lots of effort expended suggest a break
• Too easy tasks chosen engage in talk about fear
  of failure

57
Functional Components

• Set of connections between events (inputs) and
  key components of the model e.g. its nodes and
  links.
• Mechanism that changes the state of the nodes
  following some updated input.
• Sets of rules that embody the motivational
  pedagogical theory. That is, two theories that
  link actions, outputs or reactions of the system
  to expected changes in the internal cognitive and
  affective state of the student.

58
TWO KINDS OF THEORY

• How systems should act or react in order to
  change the motivational state of the student e.g.
  Keller, ARCS model
• How exactly a students motivational state
  affects her learning e.g. if good state of
  motivation then good learning occurs

59
CONTINUUM OF AFFECTIVE MODELS

• Fully articulated emotion models e.g. OCC
• Models of educationally relevant affective
  dimensions
• Process models e.g. bayesian (Conati)
• Simple thermostat variables e.g. Attention,
  relevance, confidence, satisfaction, curiosity,
  independence, control (Keller ARCS)

60
(No Transcript)
61
What MORE did when a learner asked for help
62
Pilot evaluation of MORE

• Compared versions with and without motivational
  rules
• Motivational version generally liked except e.g.
  its occasional refusal to
• provide help
• allow student to quit from problem
• Plausibility

63
CONCLUSIONS

• Multi-dimension framework for motivationally
  intelligent systems including meta-affective
• Two kinds of theory
• How systems should act or react in order to
  change the motivational state of the student
• How exactly a students motivational state
  affects her learning e.g. if good state of
  motivation then good learning occurs
• Attention to motivational issues
• Some performance gains
• Some process differences
• Interactions with ability

64
FURTHER WORK

• Develop and apply motivational framework to
  further systems
• Explore domain independent tactics e.g. breathing
  exercises, mantras
• Explore meta-affective level

65
Thanks to the work of

• Rose Luckin
• Teresa del Soldato
• Genaro Rebolledo Mendez
• Erika Martinez-Miron
• and others in the HCT Group

66
(No Transcript)
67
Activity context
Learning
Comparison