Peter Kvam
New theories and models of cognition are often limited by their tractability – how easy they are to fit, compare, construct, and test. Recently-developed tools in machine learning are beginning to allow us to overcome many of these computational barriers, making it possible to approach new models and applications that were previously intractable. In this talk, I review several new avenues of research that we can pursue using machine learning tools to fit, compare, and discover models of cognition. These include real-time parameter estimation, continuous-outcome responses, model fitting and comparison in big data, and joint models of multiple outcomes like visual fixations, response time, and accuracy. I will focus on a few illustrative results that show how overcoming modeling barriers can advance our fundamental understanding of the cognitive processes underlying decision-making. The talk will conclude with a forward-looking perspective on a few key innovations in modeling tools and how they might advance the fields of psychology and cognitive science more broadly. Maximizing the impact of these tools will require democratizing access to computational modeling, and I outline ways that mathematical psychologists might contribute to such an effort.
This is an in-person presentation on July 26, 2025
(16:00 ~
17:30 EDT).
