Calendar of events

Title: Rethinking parameter invariance for portability and compatibility in cognitive modeling.

Abstract: A core goal in cognitive and psychological sciences is developing theories and models that have the capacity to generalize (e.g., from the laboratory to the real world). However, meeting this benchmark is difficult for several reasons, such as, 1) human behaviour is complex and can vary as a function of context and individual differences, 2) our measures of psychological constructs are approximate and noisy, and 3) we lack standardized methods for testing for invariance. The current talk focuses on the latter issue in the context of cognitive modelling by exploring the relative tradeoffs of two major approaches -- joint task fitting and across-task parameter substitution. Using established results, simulations and proofs I compare the relative strengths and weakness of each approach, and provide recommendations on their application.