Currently, I am a graduate student with Michael D. Lee in the Department of Cognitive Sciences at the University of California, Irvine.
My research interests are fairly broad, encompassing large swaths of behavior, neuroscience, and the application of statistical and machine learning techniques to analyzing data from experiments. Much of what I do would be considered “modeling”. Typically, it consists of either developing a new formal theory of performance on a psychological task or comparing existing theories of a task, to illuminate aspects of how people are performing it. Up to now, my work has been restricted to feature salience in stimulus representation, statistical models of individual differences, decision-making and problem-solving, and Item Response Theory (IRT). In future work, I would like to extend into more areas of behavior, such as learning and memory, as well as exploring the neural underpinnings of behavior and statistical models linking behavioral and imaging data.
Feature Salience
My dissertation work deals with developing methods for finding which features of stimuli people find salient as well as a theory of why people find the more salient features are more salient. My work on finding the features people find salient consists of modeling the relationship between feature generation tasks, in which participants generate lists of features for each stimulus in a collection, and similarity judgment tasks, in which participants judge the similarity between each pair of stimuli in a collection, in order to determine how important each of the generated features is to representing the similarities between stimuli. On understanding why some features are more salient than others, previous work has compared a number of simple heuristics for choosing salient features in order to illustrate which qualities of the relation between a feature and a category predict whether the feature will be salient. Current work deals with why certain features become strongly associated with a particular category, treating the associations as the result of an inference over categories and salience from observed exemplars of the a priori unknown categories.
Individual Differences
In addition to feature salience, I have looked at latent variable methods for accounting for variations in task performance across individuals. Thus far, I have explored two ways in which individuals can vary. In the first, all individuals are assumed to perform the task according to a single parameterized model, with variation arising from their individual model parameters. In a model of individual differences in bandit problem performance, variation was modeled as emerging hierarchically from group level in idiosyncratic factors. Alternatively, I have looked at the case where different groups of individuals perform the task in fundamentally different ways. One such situation is when some individuals cheat on a task, as when a participant always chooses the same response on a forced choice task. Here one group of individuals is assumed to honestly perform the task and one or more groups of individuals are assumed to be cheating and which individuals belong to which groups are inferred from the data. A future collaboration with Dr. W. Rodman Shankle, a clinician, will look at using this method to find cheaters within a telephone administration of the Mild Cognitive Impairment (MCI) screening test for Alzheimer's patients.
Item Response Theory
The final component of my work involves extensions and applications of IRT. In collaboration with Eric-Jan Wagenmakers at the University of Amsterdam, I am looking at applying IRT to extend an existing model of interpersonal agreement. With William Batchelder and Mark Steyvers, I am developing an extension of IRT to matching tests, in which participants are shown two lists of equal length and asked pair items of the first list in items of the second in such a way that each item in the second list is paired with an item in the first and vice versa.