1. Self-Adaptive Learning through Teaching
Evgeny Karataev
School of Information Science
University of Pittsburgh
(advisor: Vladimir Zadorozhny)
Problem: adaptive information processing in a large scale decentralized and
loosely coordinated systems based on crowdsourcing and social computing.
We propose to build an adaptive on-line social network to improve the
process of learning.
Therefore, in addition to the research contribution, our project will engage
students in an active learning through teaching process.
Abstract
Ways of teaching and
ways of learning vary greatly.
Challenge: adaptive large scale information processing.
Motivation
Approach
Complex Adaptive Information System (CAIS)
Ease of sharing
and getting information
Large amount
of data
Overload
“one size does not fit all”
• Design – to utilize crowdsourcing techniques, collaborative filtering and
collective intelligence
• Monitor – dynamic data warehousing, data analysis and data mining
techniques
• Adapt – adaptively converge to the most productive learning pathways
with respect to a particular group of students and their
performance profiles
Research Questions
• An agent-based simulation, to study behavior of the SALT:
• Agent:
• Student (with randomly generated parameters)
• Agent’s Rules:
• Create lesslet (based on agent parameters)
• Take lesslet (based on lesslet and agent parameter)
• Adaptive environment:
• Recommend lesslet
By controlling input parameters and experimenting with new algorithms we can
study various aspects of CAISs such as their convergence properties,
their sensitivity to initial conditions, and etc.
Simulation-based study
Benchmark system. SALT
Preliminary results
• Study of simulation model to analyze and find trends and patterns, as well
as phase transition parameters.
• New adaptive algorithms to recommend lesslets, learning pathways or
friends.
• Large graph visualization to allow easy, interactive and adaptive way of
exploring topics and lesslets, as well as learning pathways and users
activities.
Future work
class # of
students
avg
success
#of best
pathways
est avg
successa
UG 34 79.54% 30 97.64%
G 31 81.27% 27 95.63%
aaverage success of students if each of them
follow the personal best possible pathway
• Self-Adaptive Learning through Teaching (SALT) – to gain educational
knowledge via adaptive learning through teaching.
Adaptive
Environment
Agents
Rules
Global structure or patterns without a central authority or
external element imposing it through planning
Learning pathways
Best pathway