How robust are probabilistic models of higher-level cognition?

Area of Science:

• Cognitive Science
• Computational Neuroscience
• Philosophy of Mind

Background:

• The theory of the mind as a rational engine of probabilistic inference is increasingly popular.
• This perspective is applied across diverse domains like word learning, pragmatics, naive physics, and future prediction.

Purpose of the Study:

• To critically evaluate the Bayesian models of inference as a theory of mind.
• To challenge the widespread assumption that probabilistic inference equates to rationality or optimality.

Main Methods:

• Reevaluation of post hoc practices commonly used in Bayesian modeling.
• Conceptual analysis of the relationship between probabilistic inference and rationality.

Main Results:

• The Bayesian models of inference are argued to be less promising than widely believed.
• Post hoc practices associated with these models require significant reevaluation.
• The common equation between probabilistic inference and rationality/optimality is demonstrated to be unjustified.

Conclusions:

• The prevailing view of the mind as a Bayesian inference engine warrants skepticism.
• The methodological and conceptual foundations of applying Bayesian models to cognition need further scrutiny.
• A critical reassessment of the rationality assumption in cognitive science is necessary.