Peter Kvam, The Ohio State University

Michael S. Goodman ’74 Memorial Colloquium Series.  Speaker: Peter Kvam, The Ohio State University.   Title: Connecting decision data through cognitive models.   Abstract: Cognitive models and the theories they embody allow us to connect behavioral, self-report, and even neural data to sets of underlying cognitive processes. Although these models are frequently applied to a single task or type of observation, data can be also connected across tasks and measures by tying model parameters to a common set of latent traits or processes. To do so, I examine a model structure where performance on two different decision measures is linked using a “joint” model that accounts for individual differences and group-level behavior across multiple sources of data. This joint modeling approach is applied to two clinically diagnostic behavioral measures, the delay discounting and Cambridge gambling tasks, in order to determine whether they measure the same propensities for impulsive decision making. Using a model comparison method based on Bayes factors that allows us to obtain evidence for the “null” hypothesis, I show that the two tasks measure unique dimensions of impulsivity that are related to substance use and addiction. I conclude by examining additional applications of the joint modeling approach as well as other future research directions.