Abstract

Perception, according to advocates of the predictive processing (PP) framework in cognitive science, is a kind of controlled hallucination. Philosophers interested in PP, however, differ on how best to interpret this slogan. Does it suggest a new kind of idealism about perceptual objects or is it just a useful metaphor, illustrating something about how PP systems work without entailing a radical shakeup of mainstream realist views in the philosophy of perception? In this paper, I take a historically informed approach to this question, drawing on the Kantian roots of the contemporary framework. What perception reveals, according to PP, is a world shaped by the neurocomputational capacities and limitations of perceiving creatures. This means that the properties of perceived objects are, in an important and surprising sense, perceiver-dependent. Nonetheless, thanks to the integration of perceptual and agential capacities envisioned by PP, we should also accept a central claim of realism: in encountering this world we also encounter—come to be acquainted with—mind-independent particulars. PP thus supports a promising next step in the development of a Kantian argument linking the unity of agency to the objectivity of perception. The argument was advanced in its modern form by P.F. Strawson and has later been developed by Susan Hurley and others. In the revised form presented here, it says that PP-perceivers must be practically self-aware agents and that practically self-aware agents embedded in an environment can be acquainted with objects it contains.