Brendan Fleig-Goldstein (University of Pittsburgh)
Why might a scientist want to establish a cognitive model as rational or “optimal” in some sense (e.g., relative to some normal environment)? In this presentation, I argue that one motivation for finding optimal cognitive models is to facilitate a particular strategy for marshalling evidence for cognitive theories. This claim stands in contrast to previous thinking about the role of optimality claims in cognitive modelling. Previous thinking has generally suggested that optimality claims either: serve to help provide teleological explanations (explanatory role); heuristically aid in the search for predictively accurate models (methodological role); or are themselves hypotheses in need of testing (empirical role). The idea that optimality claims can play a role in the process of testing theories of cognition has not previously been explored.
The evidential strategy proceeds as follows: a scientist proposes an optimal model, and then uses this optimal model to uncover systematic discrepancies between idealized human behavior and observed human behavior. The emergence of discrepancies with a clear signature leads to the discovery of previously unknown details about human cognition (e.g., computational resource costs) that explain the discrepancy. The incorporation of these details into models then gives rise to new idealized models that factor in these details. New discrepancies emerge, and the process repeats itself in an iterative fashion. Successful iterations of this process results in tighter agreement between theory and observation. I draw upon George E. Smith’s analysis of evidence in Newtonian gravity research (e.g., 2014) to explain how this process of iteratively uncovering “details that make a difference” to the cognitive system constitutes a specific logic of theory-testing. I discuss Thomas Icard’s (e.g, 2018,) work on bounded rational analysis as an illustration of this process in action.
