Edm2015presentation

Students changing their answers, based on what

their friends say

Sameer Bhatnagar 1 Michel Desmarais 1

C. Whittaker 2 N. Lasry 3 M. Dugdale 3 K. Lenton 4 E. Charles 2

1Polytechnique Montreal

2Dawson College

3John Abbott College

4Vanier College

June 25, 2015

S.Bhatnagar, M. Desmarais (Polytechnique Montreal) DALITE 1 / 22

What is this all about?

Introducing a new data set

Demonstrating its potential for EDM


Outline

1 Glossary

2 Data

3 ResultsAnswer ChangesGender DifferencesPeer Voting

4 Future Work


Glossary

Peer InstructionClassroom Activity popularized by Eric Mazur of Harvard

DALITEWeb Based Learning Environment for Peer Instruction at HomeLet’s give the system a try!


http://edu8.ca

Outline

1 Glossary

2 Data


4 Future Work


The Data Set: from the classroom

118 students from three different colleges

Four different teachers, five different groups

Final Grade for the course (Freshman Year Physics)


The Data Set: from DALITE

7100 Student-item pairs

Each student item pair includes

First AnswerRationaleSecond AnswerHow many votes the rationale received


Outline

1 Glossary

2 Data


4 Future Work


Strong students as likely to go the wrong way

0.0

0.1

0.2

0.3

0.4

bottom topFinal grade partitioned at median

P(R

ight

−−

> W

rong

)


Weak students as likely to go the right way

0.0

0.2

0.4

0.6

0.8


P(W

rong

−−

> R

ight

)


Don’t Use the Tool Alone!

Which group never talked about DALITE in class?

0.0

0.2

0.4

0.6

cw09 cw10 KJL MD NLDifferent Groups

P(S

witc

hing

Ans

wer

)


Gender Gap in Physics

0.4

0.6

0.8

1.0

f mGender

P(R

ight

Ans

wer

on

FIR

ST

Atte

mpt

)


Gender Gap in Physics except if you let them

change their minds

0.4

0.6

0.8

1.0

f mGender

P(R

ight

Ans

wer

on

SE

CO

ND

atte

mpt

)


No Gender Gap in Votes Earned

0.0

0.2

0.4

0.6

f mGender

Ave

rage

Num

ber

of V

otes

Ear

ned


Strong students do earn more votes

0.0

0.1

0.2

0.3


Ave

rage

Vot

es e

arne

d ov

er th

e se

mes

ter


Strong, even when Wrong

0.0

0.2

0.4

0.6


Ave

rage

Vot

es e

arne

d ov

er th

e se

mes

ter

for

WR

ON

G A

NS

WE

R


Outline

1 Glossary

2 Data


4 Future Work


Future Work

NLP on Rationales

Automated Essay ScoringTopic Modeling

Collaborative Filtering with voting data


Welcome to The Big Leagues

License to HarvardXNext Step: Your edX MOOC!


http://studio.edge.edx.org/

What is this all about?

Introducing a new data set

Demonstrating its potential for EDM


Acknowledgements

Michel Desmarais

SALTISE researchgroup

PAREA


Thank you!

Any Questions?

Contact Infowebsite: sameerbhatnagar.github.io


http://sameerbhatnagar.github.io

Edm2015presentation

Data & Analytics

Transcript of Edm2015presentation