Ph.D. Defense of Dissertation: Maithilee Kunda

Add to Calendar
Date:
December 17, 2012 1:00 pm - 3:30 pm
Location:
TSRB TBD - contact student

Ph.D. Defense of Dissertation Announcement
Title: Visual Problem Solving in Autism, Psychometrics, and AI: The Case of the Raven's Progressive Matrices Intelligence Test

Maithilee Kunda
Computer Science Ph.D. Student
School of Interactive Computing

Date: Monday, December 17th, 2012
Time: 1:00PM - 3:30PM EDT
Location: TSRB (Room TBD)

Committee:

  • Dr. Ashok K. Goel (Advisor, School of Interactive Computing, Georgia Tech)
  • Dr. Gregory Abowd (School of Interactive Computing, Georgia Tech)
  • Dr. Nancy Nersessian (School of Interactive Computing, Georgia Tech)
  • Dr. Brian Scassellati (Department of Computer Science, Yale University)
  • Dr. Eric Schumacher (School of Psychology, Georgia Tech)
  • Dr. Isabelle Soulieres (Department of Psychology, Universite du Quebec a Montreal)

Abstract:
Much of cognitive science research and almost all of AI research into problem solving has focused on the use of verbal or propositional representations.  
However, there is significant evidence that humans solve problems using multiple different representational modalities and strategies, including visual or iconic ones.  In this thesis, I investigate visual problem solving from the perspectives of cognition in autism, psychometrics, and AI.

Studies of individuals on the autism spectrum show that they often use atypical patterns of cognition, and anecdotal reports have frequently mentioned a tendency to "think visually." In my thesis, I provide a precise characterization of visual thinking in terms of iconic representations.  I conducted a comprehensive review of data on several cognitive tasks from the autism literature and found numerous instances indicating that some individuals with autism may have a disposition towards visual thinking.

One task, the Raven's Progressive Matrices test, is of particular interest to the field of psychometrics, as it represents one of the single best measures of general intelligence that has yet been developed.  Typically developing individuals are thought to solve the Raven's test using largely verbal strategies, especially on the more difficult subsets of test problems.  In line with this view, computational models of information processing on the Raven’s test have focused exclusively on propositional representations.  However, behavioral and fMRI studies of individuals with autism suggest that these individuals may use instead a predominantly visual strategy across most or all test problems.

To examine visual problem solving on the Raven's test, I first constructed a computational model that uses a combination of affine transformations and set operations to solve Raven's problems using purely pixel-based representations of problem inputs, without any propositional encoding.  I then performed four different analyses using this model.  

First, I tested the model against three versions of the Raven's test, in order to determine the sufficiency of visual representations for solving this type of problem.  
I found that visual representations of this form can successfully enable solving 50 of the 60 problems on the Standard Progressive Matrices (SPM) version of the test, which is comparable to the performance of the best computational models that use propositional representations.  Second, I evaluated model robustness in the face of changes to its representation of pixels and of visual similarity.  I found that varying these low-level representational commitments causes only small changes in overall model performance.  Third, I performed successive ablations of the model to create a new, data-based classification of problem types on the SPM, based on which affine and set transformations are necessary and sufficient for finding the correct answer.  Fourth, I examined whether patterns of errors made on the SPM can provide a window into whether a visual or verbal strategy is being used.  While many of the observed error patterns were predicted by considering aspects of the model and of previous human studies, I found that error patterns overall do not seem to provide a clear indicator of strategy type.

The main contributions of this dissertation include: (1) a rigorous definition and examination of a disposition towards visual thinking in autism across several different cognitive tasks; (2) a sufficiency proof, through the construction of a novel computational model, that visual representations can successfully solve many Raven's Progressive Matrices problems; (3) a new, data-based classification of problem types on the SPM; (4) a new classification of conceptual error types on the SPM; and (5) a methodology for analyzing, and an analysis of, error patterns made by humans and computational models on the SPM.  More broadly, this thesis is a contribution towards our understanding of visual problem solving.