Australasian Mathematical Psychology Conference 2019

Time-varying cognitive models of decision making

Guy Hawkins
Psychology, University of Newcastle
David Gunawan
University of New South Wales
Robert Kohn
University of New South Wales
Scott Brown
University of New South Wales

Almost all cognitive process models of decision making assume that the latent parameters driving performance are stationary across trials. This conflicts with intuition, and data, that performance does not change with increasing exposure to a task. Here, we outline a flexible hierarchical Bayesian framework that allows for across-trial dynamics in the parameters of decision making models, and thus makes time-varying predictions for behaviour. We demonstrate the approach with the Linear Ballistic Accumulator (LBA) model. We show that time-varying LBA models reliably recover the data-generating model in simulated data, and they are consistently selected over equivalently specified stationary LBA models. Furthermore, the time-varying LBA model provides a good account of across-trial dynamics observed in choice and response time data, and time-varying parameter estimates that provide insight into the dynamics of latent cognitive mechanisms driving observed decision behaviour.