Kolloquium

Mittwoch, den 04.02.2026, um 14:15 Uhr trägt Florian Seitz (Uni Basel) bei uns im Seminarraum (03-013), in Präsenz vor.

Title: From Similarity to Rules and Back Again: Unifying Principles of Generalization and Inference

Abstract: Human inference relies on the ability to generalize beyond observed information. Theories of cognition often distinguish between similarity-based processes, which extrapolate from stored exemplars, and rule-based processes, which apply abstracted principles, treating them as competing explanations of inference. In this talk, I challenge this sharp division. First, I introduce similarity-based and rule-based generalization as core processes underlying categorization, judgment, and inference more generally. I then revisit the category variability effect, commonly interpreted as evidence against similarity-based accounts, and show how a novel model using an alternative similarity measure can capture this effect—provided the effect itself proves to be empirically reliable. Next, I address the broader challenge of distinguishing similarity-based from rule-based processes. I present evidence from an eye-tracking study demonstrating that the inference process people rely on according to cognitive modeling is reflected in their eye movements. Finally, I introduce a new extrapolating exemplar model that reunites rule-like generalization patterns with similarity-based processing, offering a unified framework in which rule- and similarity-based behavior can emerge from shared principles rather than competing systems.