New publication in "Psychological Review"

6 January 2025, by Uğur Turhan

To provide a mechanistic explanation for observed behaviours, increasingly more psychologists rely on cognitive computational models which map the observed behaviour onto multiple model parameters, each representing a latent psychological variable. In this article, we take a critical look at how such cognitive models are being currently used and we argue that the dominant approach fails to capture important questions - namely, how or why the individual settled on a particular behaviour. We suggest that, rather than being viewed as constant, unchanging states which generate behaviour, model parameters are more accurately represented as dynamic cognitive states which change over the course of a task in response to the participant's goals, their capabilities and the informativeness of the environment. We provide specific suggestions for researchers on how to account for these factors when creating their models and the task, and we review several approaches which allow for a more dynamic estimation of evolving parameter values. Ultimately, our article aims to bring attention to the ways in which the already-powerful method of cognitive modelling could be leveraged even further, in order to help us understand more facets and nuances of human behaviour.

Link to the article