The slides from my defense of my master thesis where I applied artificial intelligence techniques to automatic selection of sequences of Open Educational Resources based on the measured impact they have on learning. I used Genetic Algorithms and UCB-1 selection, it was evaluated in the setting of an online course. Contact me if you want to read the thesis. The defense was successful.
1. Open Survival of the fittest
Open in the jungle of
Open Educational Resources
15th of August 2014
a Master Thesis by
Sander Latour
2. Open Educational Resources [1]
Learning objects that can freely be
reused, revised, remixed and redistributed.
[1] Daniel E Atkins, John S Brown and Allen L Hammond. Creative Common, 2007. A review of the
open educational resources (OER) movement: Achievements, challenges, and new opportunities.
3. Open Educational Resources
Learning objects that can freely be
reused, revised, remixed and redistributed.
Textual objects Video objects Interactive objects
You should not focus
on every detail. Stick
to the bigger picture
Example:
You are reading this.
Stick to the
bigger picture
14. Survival of
the fittest
UCB +
a Genetic
Algorithm
[2] A.E. Eiben and J.E. Smith. Natural Computing,
2007. Introduction to Evolutionary Computing.
[2]
15. ( )
2 ln(n)
n
NLG
average
total nr. of
evaluations
nr. of times
tried
UCB-1[3]
[3] P. Auer, N. Cesa-Bianchi and P. Fischer. Machine learning, 2002.
Finite-time analysis of the multiarmed bandit problem.
16. The impacts of these sequences are not independent
If these two
are effective
then it makes
sense to try this
17. Genetic Algorithms
Population containing a subset of candidates
Candidates have a “fitness” value, i.e. how good is it?
Higher fitness means higher chance of reproduction
Produced offspring is a combination of both parents
Inspired by Darwinian evolution
18. ( ) 2 ln(n)
n
NLG
Current population
T1
T2
Selecting most
promising sequence
Evaluation of impact
19. NLG1
NLG3
NLG2
NLG4
NLG5
Roulette selection of parents
1
2
3
4
Crossover
& Mutation
Crossover
& Mutation
Offspring
Offspring
Generational replacement
with elite preservation
elite
offspring
Current
generation
Next
generation
22. Nim game
Curriculum
Low High
student groups
4lessons
4OER
T1
T2
7sequences in
1 generation
10evaluations in
1 generation
2elite members
5%mutation
237total usable
participants
Algorithm Participants
voluntary
participation
could stop at
any moment
diverse crowd
not just students
3MC 3MC
23. Does the system learn to pick sequences with
more learning impact over those with less impact?
Figure: Regret in Rules - Low Figure: Regret in Intuition - Low
Built-up
regret
students
The system worked for the “Low” student groups
In “High” groups there was either too little data or a technical issue
It’s unknown how good the apparently best sequences really are
24. Possible explanation
limited pre- and post-test
coarse division of students
independence assumption
Variance in the observed learning impact
learning
impact
students
Figure: Best sequence in “Rules” lesson for student group “Low”
25. In conclusion,
A possible approach for using learning
impact in the assessment of OER
has been presented and tested.
Many lessons can be drawn from the
results, but the principle works.
I recommend others to continue
on the path of using learning impact
in the assessment of OER.
27. Diversity of individuals in each population
Diversity was low with few novelties
Parameters were set to converge more quickly
Exploration was often not possible by means of crossover