My research is broadly aimed at developing computational models of social cognition. I am interested in how people can infer beliefs and intentions in others, by observing others’ actions and employing recursive theory-of-mind (e.g. inverse reinforcement learning, social psychology). I am also interested in group-level equilibria when agents are collaborating or competing (e.g. game theory, agent-based modeling). Finally, I am interested in mechanism design and other ways quantitative characterizations of a phenomenon can be used to predict and shape behavior.
Partial Abstract: We are constantly barraged with information, and yet to navigate the world successfully we must be capable of isolating relevant units. The process of learning statistical relationships between items or events is called statistical learning, and it is a ubiquitous and powerful learning mechanism spanning many modalities and domains. With such a powerful learning mechanism at their disposal, humans must use strategies to direct their abilities in effective manner. We sought to understand how people use two strategies, matching and maximization (loosely analogous to exploration and exploitation respectively), to learn sequences that mimic the complexity of the real-world visual environment. Subjects observed sequences of symbols generated by a first-order Markov process and were asked to predict the upcoming stimulus. Computational modelling was used to extract the learning profiles of participants and the strategies that they followed in learning patterns and making predictions. Using these indices, we tested subjects in seven different conditions, hypothesizing that subjects would change their strategies based on different situational demands.
Abstract: Trichromatic color vision in humans and macaque monkeys depends upon the activity of three classes of cone photoreceptors in the retina. The brain computes a representation of color by comparing the activity of these cone types through multiple stages of neural circuits, although the mechanisms are not well understood. One region thought to be important is V4/Posterior Inferior Temporal Cortex (PIT), a mid-tier stage in the visual-processing hierarchy that contains color-selective cells spatially clustered within “globs” identified by fMRI. Conway, Moeller, and Tsao (2007) used fMRI-guided single-unit electrophysiology to record from these cells in fixating monkeys, measuring the cells’ color tuning as a function of luminance changes. This physiology experiment was replicated psychophysically in humans to test the hypothesis that these neurons are correlated with color perception. We hypothesize that the glob cells represent the neural correlate of the perceptual Bezold-Brücke effect, a subtle color change that can be introduced by modulating luminance (a light appearing green at high luminance requires more long-wavelength power to appear the same green at low luminance). Cell responses to 45 colors at three different luminance levels (low, equiluminant, and high, relative to the adapting background) were analyzed, and human subjects were required to perceptually match the equiluminant and high luminance colors. Cell color preferences were largely stable against luminance changes (“luminance-invariant”), but some cells showed prominent color preference shifts across luminance levels. Human perceptual color matching revealed consistent luminance- invariant responses across luminance levels. Neither the tuning properties of macaque glob cells nor the perceptual color matching experiments in humans supported the predicted Bezold-Brücke effect. Moreover, the tuning properties of the cells did not correspond to the luminance-invariant perceptual responses. These results do not support, but also do not preclude that glob cells are implicated in color perception. Future studies will contain more finely sampled stimuli and cell responses from areas downstream from V4/PIT, including Anterior Inferior Temporal Cortex (AIT).