Proceedings of the Annual Meeting of the Cognitive Science Society

Active Inference and its accompanying Bayesian Mechanics (BM) are important psychological and cognitive science theories. While there is a strong interaction between the theories and the philosophical realm, it needs to be clarified what its metaphysical commitments are. We tease out these commitments while looking from the perspective of the psychology of goals. We find that Active Inference cannot account for the dynamic growth of goals, primarily because of its closed generative model. We trace the reason for this through the extrinsic `ontological constraint' of BM, characteristic of all mechanistic models which follow the `logic of machines.' Finally, we ground our arguments in the necessity of external relations in substance metaphysics and its incompatibility with internal relations and impredicativity. Thus we argue that Active Inference implicitly presupposes a Substance metaphysics, yielding the theory no resource to model novelty, growth, and development observed in human psychology. We briefly sketch a powerful alternative grounded in process metaphysics to model biological and cognitive systems.