Loading presentation...

Present Remotely

Send the link below via email or IM

Copy

Present to your audience

Start remote presentation

  • Invited audience members will follow you as you navigate and present
  • People invited to a presentation do not need a Prezi account
  • This link expires 10 minutes after you close the presentation
  • A maximum of 30 users can follow your presentation
  • Learn more about this feature in our knowledge base article

Do you really want to delete this prezi?

Neither you, nor the coeditors you shared it with will be able to recover it again.

DeleteCancel

Make your likes visible on Facebook?

Connect your Facebook account to Prezi and let your likes appear on your timeline.
You can change this under Settings & Account at any time.

No, thanks

Job Talk - Groningen

Job Talk
by

Philip Pavlik

on 11 August 2016

Comments (0)

Please log in to add your comment.

Report abuse

Transcript of Job Talk - Groningen

Applying Cognitive Models to Improve Education and Human Performance
Dr. Philip I. Pavlik Jr
Carnegie Mellon University
Human Computer Interaction Institute
4th of March 2010
ACT-R
Bachelors in Economics at University of Michigan
Optimality
EEG
FaCT (Fact And Concept Training) System
Paper submitted to Intelligent Tutoring Systems 2010
Argue that theoretical synthesis is possible, even necessary
Integration may be attained by focusing on perspectives at grainsizes
Look for conflicts
No conflict, no problem
Conflict, assistance dilemma or falsification
Sometimes combinations may by synergistic















Cognitive psychology might not transfer to education often removes confounds that may be crucial interactive factors when considering their effect in educational contexts
E.g. cognitive psychology SUBJECTS always motivated to participate
STUDENTS need corresponding motivation and integrating motivation theory provides a way to integrate this fundamental interaction
Only after integrating these fundamental interatcions will principles transfer well from Cognitive Science to Education
Theoretical Synthesis
Background
First had to build a model (Pavlik & Anderson, 2005)
Japanese paired-associates experiment (Romaji) 4x3x2 design (practices, spacing intervals, retention intervals)
Decay-based model of spaing
Also made autoassociator neural network model of process (PDP class semester project)
Chinese
IES Project
Educational Datamining
Spacing Effect
Test and Study Effects
Motivation and Metacognition
Application as the driver
Came to realize that FaCT System was not close enough to user
Project to integrate with the Social Platform Elgg
First stages of completion
Elgg
Assistance Dilemma
Perceptual Design
Microeconomics
Game Theory
Never was exposed to Cognitive Modeling


Seemed to describe an Economics of the mind
Also corresponded to my interest in learning science
Took several Psychology course at local university
Became interested in mental rotation
Was accepted to work with John Anderson on mental rotation models
Switched to memory modeling work after year 1 because of my interest in declarative learning
Key revelation!
Cognitive Models can be be optimized to discover how to optimize the systems they describe
Optimizing
Needed to better understand the difference between test and study practice
Better undestand the effect of review after success
Better understand the effect of duration of study
Better parameters for the model (ACT-R had/has no model of study duration that makes sense)
Also looked at effect of mnemonic strategy
Importantly-- Compute optimium study duration
Application in Lab
180 word pairs
Japanese-> English
3 one-hour learning sessions
testing session (2 cycles through set) after 1 week
2 Control conditions
Postdoctoral work at PSLC (Pittsburgh Science of Learning Center)
Funded as research by local philanthropist
Developed ideas from graduate work
Flexible System
Sequence of Chinese Classroom Experiments
Build an understanding of the classroom use of these systems
Difficult to produce clear long-term learning benefits in classroom
Vastly different ways students use the tutor (much more prior knowledge)
One primary problem with long-term testing (particulary with my motivated population) is with compensatory equalization
Bridging the Bridge to Algebra (BTA) Cognitive Tutor
4 year US Department of Education Grant
Currently on Year 3
IES - Institue of Educational Sciences

Year 1 - Modeling BTA + Transfer item creation
Year 2 - Classroom Experiment and continued modeling
Year 3 - Classroom Experiment and continued modeling
Year 4 - Theoretical reassessment and integration
Goal was to find what simple skill components (e.g. chunks or productions) that are necessary for effective progress in-pre-algebra
Method was to look at pairwise comparisons
Results from IES Classroom experiment 1 showed many correlations between skill components, but no true transfer
Problems: perhaps grainsize isn't really as fine as we were looking OR perhaps carryover effects from multiple components canceled out detection of effect
Transfer
Failure of transfer in year 2 experiment corrected for year 3 experiment
Now using stronger control (no treatement)
Now also using stronger manipulation
Inference trials
Analogy trials
Provide a perceptual representation that supports an extensive mental model -- This is an example of theoretical synthesis of design, perceptual learning and situated cognition principles
From a cognitive psychology perspective this means provide a clear problem schema template
In ACT-R terms this is a flexible collection of chunks that are well learned with associated production rules that have generalized conditions (left hand sides)
It seems that the sum of these perspectives suggests a new approach to education software tutor design
Work with Ken Koedinger
Development of "Assistance Dilemma" theory in other domains
Assistance dilemma -- supposes that many tradeoffs in learning can be characterized as more or less assistance
E.g. less spacing more assistance
Paper in preparation will present how this applies to the worked-example (more assistance) and problem solving (less assistance) situations
New funding to look at metacognitive and motivational variables in Chinese Courses
Likert scale measures
MSLQ Motivated Strategies and Learning Questionnaire
Vocabulary memory strategy assessment
Pre and post quiz
See how these relate to FaCT System variables
Small side project
Purchased a Neurosky EEG device
Plan to build a classifier Expectation:
Covariance Search
Pairwise Search
Problems with lack of synthesis in educational application
Perspective by Grainsize Analysis
Assistance Dilemma Work Drives Memory Model Development
Structural Equation Modeling
Combining variables from different domains
Global opimality given some fixed inputs, some modifiable inputs, and some outputs
Compute optimum settings for modifiable inputs to maximize outputs
This may present a framework to resolve questions of optimality that cross disciplines
3 seminars
Application
Genc Doko
Lili Wu
Giancarlo Dozzi

Sushanth Tiruvaipati
James Leszczenski

Jordan Bartlow
Dave Hora
Chris Broglie

Programmers
Part-time Programmers
Postdoc and Grad Students
Student Research Assistants
Part-time student programmers
Jozsef Toth, post doc
Hao Cen, doctoral student
Nora Presson, doctoral student

Eric Emerson
Thomas Bolster
Arthur Tu
Suhrob Munavarov

Ken Koedinger
Brian MacWhinney
John Anderson

Executive Guidance
Team
A failure that is followed by a success will have a different pattern (alpha, beta, gamma stronger) than a failure that is followed by another failure
Read John Anderson's book
http://prezi.com/agjdgio2rfns/
To review this talk
Full transcript