From Logic to Behavior

Jakub Szymanik (Amsterdam) gives a talk at the MCMP Colloquium (12 June, 2013) titled "From Logic to Behavior". Abstract: In this talk I will explore the applicability of modern logic and computation theory in cognitive science. I will show how logic can be used to build cognitive models in order to explain and predict human behavior. I will also illustrate the use of logical and computational toolboxes to evaluate (not necessarily logical) cognitive models along the following dimensions: (i) logical relationships, such as essential incompatibility or essential identity; (ii) explanatory power; (iii) computational plausibility. I will argue that logic is a general tool suited for cognitive modeling, and its role in psychology need not be restricted to the psychology of reasoning. Taking Marr's distinctions seriously I will also discuss how logical studies can improve our understanding of cognition by proposing new methodological perspectives in psychology. I will illustrate my general claims with examples of the successful research on the intersection of logic and cognitive science. I will mostly talk about two research projects I have been recently involved in: computational semantics for generalized quantifiers in natural language and logical models for higher order social cognition. The major focus will be computational complexity and its interplay with "difficulty" as experienced by subjects in cognitive science.