09/08 Peter Pirolli, PARC - Abstract
09/15 Mary Peterson, University of Arizona - Abstract
10/13 Alex Pouget, University of Rochester - Abstract
10/27 Neal Cohen, University of Illinois at Urbana-Champaign - Abstract
12/01 Gary Dell, University of Illinois at Urbana-Champaign - Abstract

Abstract

9/8: Peter Pirolli, PARC
Title: From Solo to Social Information Foraging Theory
Abstract: Information Foraging Theory is a theory of human-information interaction that aims to explain and predict how people will best shape themselves to their information environments, and how information environments can best be shaped to people. The approach involves a kind of reverse engineering in which the analyst asks (a) what is the nature of the task and information environments, (b) why is a given system a good solution to the problem, and (c) how is that “ideal” solution realized (approximated) by mechanism. Typically, the key steps in developing a model of information foraging involve: (a) a rational analysis of the task and information environment (often drawing on optimal foraging theory from biology) and (b) a computational production system model of the cognitive structure of task. I will review work on individual information seeking, as well as out more recent studies of the social production, sharing, and use of information in areas such as wikis, social tagging, social network sites, and social search.

9/15: Mary Peterson, University of Arizona
Title: Figure-Ground Perception: From Familiarity to Competition
Abstract: The determination of where objects, or figures, lie with respect to edges in the visual input is a critically important aspect of perception. When an edge is perceived as a boundary for an object on one side, the region on the other side simply looks like a local background to the object; its shape is given by distant edges rather than by the edge is shares with the figure. Progress in understanding figure-ground perception was hampered by the long-held assumptions that (1) figure-ground segregation was an early stage of processing, and (2) effects of past experience (e.g., familiarity) could not penetrate that stage. I will review research showing that past experience can affect figure-ground perception. I will then discuss recent behavioral and neurophysiological experiments indicating that rather than being a stage of processing, figure-ground perception results from cross-edge inhibitory competition between two potential objects; the two contenders are identified in a first pass of processing. The cross-edge competition entails mutual inhibition. The stronger contender is inhibited less than the weaker one; as a consequence it is perceived as the shaped figure. The weaker contender is inhibited more than the stronger one; as a consequence, it appears shapeless.

10/13: Alex Pouget, University of Rochester
Title: Bayesian decision making with probabilistic population codes
Abstract: The brain often faces the task of making perceptual decisions on the basis of uncertain sensory evidence, e.g. hitting a fast approaching tennis ball. Such decisions involve several stages, two of which are evidence accumulation and response selection. In the evidence accumulation stage, the best strategy consists in computing a probability distribution over the sensory variables given all the evidence available over time. For the response selection, the goal is to select the optimal response given the probability distribution at decision time. We show that, when neurons exhibit Poisson-like variability, both stages of the decision process can be formulated in terms of simple operations on a type of neural code we call probabilistic population codes. The accumulation of evidence simply requires a temporal integration of neural activity while response selection can be done optimally through attractor dynamics. This theory works for N-choice decisions where N can take any value, as well as for decisions over continuous variables. We show that this model can fit existing psychometric and chronometric functions, and can account for neurophysiological data in LIP. This framework also predicts that LIP encodes the confidence of the decision, a prediction which we have tested on LIP data and for which we present preliminary evidence. More generally this framework provides a way to test whether neural circuits perform optimal Bayesian inference, not just for decision making in the neocortex of mammals, but for any task in any animal.

10/27: Neal Cohen, University of Illinois at Urbana-Champaign
Title: Relational Memory and the Long Reach of the Hippocampus
Abstract: The talk will emphasize the following issues: (1) Studies of amnesia have made crucial contributions to our understanding of the functional role of the hippocampus in memory; (2) the hippocampus mediates a fundamentally relational form of memory; (3) two critical hippocampal processes illuminated by our current work are relational memory binding and relational memory (re)activation; (4) we have developed methods for testing human memory without appealing to verbal reports & conscious introspection, providing new opportunities for studying various patient populations and for enhancing the linkage of the human work with animal work; (5) the role of hippocampus extends to binding and (re)activation not only of long-term memories but also of information in working memory and even information still available to the senses; and (6) the role of hippocampus and the relational memory it supports can be shown to extend further, beyond memory, in areas of language, decision-making, and future imaginings.

12/1: Gary Dell, University of Illinois at Urbana-Champaign
Title: TBA
Abstract: TBA