26th APS Annual Convention: Mark Your Calendar (San Francisco, CA, USA - May 22-25, 2014)

Invited Symposium

Bayesian Methods in Cognition

Saturday, May 24, 2014, 11:30 AM - 12:50 PM
Yosemite Room B

Mark Steyvers Chair: Mark Steyvers
University of California, Irvine

Research in cognitive psychology faces many challenges: Data are noisy, participants are idiosyncratic, and models are disputed. As the talks in this symposium demonstrate, Bayesian methods are ideally suited to bring order to chaos by connecting separate pieces of information in a coherent manner.

 
Subject Area: Cognitive

Angela J. Yu

Learning and Decision-Making in Inhibitory Control
Angela J. Yu
University of California, San Diego
The ability to stop inappropriate behaviors is critical for everyday cognition and impaired in psychiatric conditions such as drug abuse. I present a Bayesian model that shows healthy humans continuously monitor the probabilistic need for stopping, and that this function deteriorates behaviorally and neurally with increased stimulant (e.g. cocaine) use.


Wolf Vanpaemel

Theory Testing with the Prior Predictive
Wolf Vanpaemel
KU Leuven, Belgium
A straightforward way to test a theory involves assessing whether its predictions are confirmed by empirical observations. In this talk, I explore how Bayesian methods can be used to test a model by its predictions, in a way that lets the model speak for itself.


Naomi H. Feldman

Modeling statistical learning in language acquisition
Naomi H. Feldman
University of Maryland
Children acquire their native language from the linguistic input with very little explicit instruction. I show how Bayesian models can help us understand the strategies that make them successful, focusing on learning synergies that arise from acquiring several aspects of language (e.g., sounds, words, and meaning) simultaneously, rather than sequentially.


Mark Steyvers

A Bayesian approach to jointly model Neural and Behavioral Data
Mark Steyvers
University of California, Irvine
We propose a flexible Bayesian framework for jointly analyzing neural and behavioral data that applies to a wide variety of cognitive models and neurophysiological measures. The framework extends standard two-stage model-based methods where researchers first estimate the parameters of the behavioral model and then correlate the parameters to the neural data. In our approach, the parameters of the cognitive and neural model as well as the parameters that link the two models are simulatenously estimated. One advantage is that the neural data can directly influence the parameters of the behavioral model.


Email Bookmark and Share