Education as a computational science Pierre Dillenbourg, EPFL EPFL - - PowerPoint PPT Presentation

▶
education as a computational science
SMART_READER_LITE
LIVE PREVIEW

Education as a computational science Pierre Dillenbourg, EPFL EPFL - - PowerPoint PPT Presentation

Aug 24, 2022 181 likes •637 views

Education as a computational science Pierre Dillenbourg, EPFL EPFL MOOCS: 1908876 Hype is over but MOOCs continue to grow . 1815471 Registrations 77 Courses Online 97510 Passed 35 Courses In Preparation N=1728 0.4

slide-1
SLIDE 1

Education as a computational science

Pierre Dillenbourg, EPFL

slide-2
SLIDE 2

EPFL MOOCS: 1’908’876

slide-3
SLIDE 3

1’815’471 Registrations 97‘510 Passed 77 Courses Online 35 Courses In Preparation

Hype is over but MOOCs continue to grow ….

slide-4
SLIDE 4
  • N=5295

N=5666 N=938 N=951 N=1728 N=1334 −0.2 0.0 0.2 0.4 NONE VIEW ACT

MOOC Usage EPFL Grade

Baccalaureat Level

  • HI

LO

Patrick Jermann, Francisco Pinto (EPFL CEDE)

EPFL Freshmen

slide-5
SLIDE 5

Education Technologies Learning Analytics Teaching CS

CS ED

slide-6
SLIDE 6

e-learning ?????????

Pierre Dillenbourg, EPFL

slide-7
SLIDE 7

Swarm Cellulo (Ayberk Ozgur, Wafa Johal, P. DIllenbourg)

slide-8
SLIDE 8

Education Technologies

CS ED

Swarm Interactions

slide-9
SLIDE 9

Johal, Lemaignan, Asselborn, Jacq, Billard, Paiva, Dillenbourg

slide-10
SLIDE 10

Education Technologies

CS ED

Teachable Agents Swarm Interactions

slide-11
SLIDE 11

Education Technologies

Learning Analytics

Teaching CS

CS ED

slide-12
SLIDE 12

Computational Models

Pierre Dillenbourg, EPFL

Education Research

learning Analytics

predict, classify, decide, ‘explain’ SVM, KMC, DNN, RNN, 2AM, POMDP

slide-13
SLIDE 13
slide-14
SLIDE 14

Kt Bt Observable State Hidden State Kt+1 Bt+1 p(Kt+1 | Bt+1, Kt )

Bayesian Knowledge Tracing

p(Kt = ‘skill-x’ | Bt= ‘correct answer’)= 1 - Guess p(Kt = ‘skill-X’ | Bt= ‘incorrect answer’)= 0 + Slip

slide-15
SLIDE 15
slide-16
SLIDE 16

Computational Models

Pierre Dillenbourg, EPFL

Education Research

Improve the management

  • f education systems
slide-17
SLIDE 17

What about modeling learning

  • utsider technology-based environments ?
slide-18
SLIDE 18

Pierre Dillenbourg, EPFL

Campus Analytics

pre-requisites p(Succeed (CS243) | Failed (CS201)) carreer p(Salary > T | {INF201, MA203,INF233,..}) recommender 78% of those who select CS243 also selected CS411

…….

slide-19
SLIDE 19

how deep ?

slide-20
SLIDE 20

Computational Models

Pierre Dillenbourg, EPFL

Education Research

Relevant Behavioral Abstractions (Features)

Education needs explainable AI

slide-21
SLIDE 21

gaze(a)=ƒ(gaze(b))

Relevant Behavioral Abstractions (Features)

slide-22
SLIDE 22

Gaze Recurrence

slide-23
SLIDE 23
slide-24
SLIDE 24
slide-25
SLIDE 25

gaze(listener)=ƒ(gaze(speaker))

Relevant Behavioral Abstractions

Feature: Gaze recurrence Context: Collaborative learning

slide-26
SLIDE 26
slide-27
SLIDE 27

gaze (learner) = ƒ (reference (teacher))

Feature: Withmeness Context: Lecturing

Relevant Behavioral Abstractions

slide-28
SLIDE 28
slide-29
SLIDE 29

Sarah d’Angelo, Kshitij Sharma, Darren Gergle, Pierre Dillenbourg (2016)

slide-30
SLIDE 30

Do finger-based or gaze-based deictics enhance learning ?

Sarah d’Angelo, Kshitij Sharma, Darren Gergle, Pierre Dillenbourg (2016)

slide-31
SLIDE 31

gaze (learner) = ƒ (gaze (teacher))

Relevant Behavioral Abstractions

Feature: ‘Withmeness’ Context: Lecturing

slide-32
SLIDE 32

Kt Bt

Modeling in the wild ?

Raca, Tormey & Dillenbourg

slide-33
SLIDE 33

gaze (learner) = ƒ (location (teacher))

Feature: Head rotations Context: Lecturing

Relevant Behavioral Abstractions

slide-34
SLIDE 34
  • L. Prieto, K. Sharma, L. Kidzinsky, P. Dillenbourg

activity (teacher) = ƒ (gaze (teacher))

slide-35
SLIDE 35

Computational Models

Pierre Dillenbourg, EPFL

Education Research

Education brings nice challenges (1) Explainability

slide-36
SLIDE 36

Education brings nice challenges (2) Cold Start

Integrate expert’s knowledge Use simulation with synthetic students

slide-37
SLIDE 37

A B

Learner 1 Learner 2 Learner 3 Learner 4 Learner 5 Learner 6 Learner 7 Learner 8 Learner 9

A A A A ?

Education brings nice challenges (3) Exploration Exploitation Tradeoff

slide-38
SLIDE 38

Explainable AI Education Context Cold Start Exploration/ Exploitation Trade-OFF Cohorte Simulations

slide-39
SLIDE 39

Multi-Armed Bandit (MAB) for Exploration-Exploitation

➢ Selecting learning activities ➢ LFA model ➢ Tested with simulated students ➢ Will present and discuss this work at ECTEL in Septembre

59

Louis Faucon, Pierre Dillenbourg, EPFL

slide-40
SLIDE 40

Education is a computational science

EPFL Center for Learning Sciences

slide-41
SLIDE 41

Happy Numbers GRAASP SpeakUP

slide-42
SLIDE 42

Co Collider Co Compu. u. Th Thinkin ing NC NCCR MO MOOCs

EPFL Digital Education Ecosystem

La Lake of Pi Piaget

Center Learning Sciences

slide-43
SLIDE 43

Education Psychology Sociology Ethnology Anthropology Economics Political Sciences Linguistics History Demography Management …..

Social Sciences

This is not one science !!!

Humanities

Modern Languages Old Language Littérature Philosophy Religion Art Musicology Museology History ….

slide-44
SLIDE 44

Education Docimology Didactics Instructional Psychology Instructional Design Learning Technologies History of Education Sociology of Education Economy of Education Special Education Psychology Cognitive psychology Social psychology Psychometry Clinical psychology Differential Psychology Developmental Psychology Sociology

And….