Navigating explanatory constraints: Predictive Processing as a hybrid mechanistic–normative explanatory framework

The next (hybrid) meeting of the seminar “Philosophy of Cognitive Science” will take place on May 15th at 10:30 AM (Warsaw time, CET), the seminar will take place in Staszic Place (IFiS, PAN), room: 161, and at GoogleMeet. Our guest will be Michał Piekarski (UKSW). We will discuss a draft titled: “Navigating explanatory constraints: Predictive Processing as a hybrid mechanistic–normative explanatory framework”

Abstract: In this paper, I propose an interpretation of Predictive Processing (PP) as a hybrid explanatory strategy that integrates mechanistic organization with a functionally motivated, normative dimension. While PP is often considered primarily as a mechanism sketch, I argue that this view does not fully account for the normative and functional dimensions of the explanations proposed by its proponents. I suggest that PP models, by identifying a range of explanatory constraints, transcend purely mechanistic and causal accounts, instead drawing on normative theories that specify not only the goals of cognitive processes but also their adaptive benefits. On this view, PP incorporates both mechanistic and functional components, the latter of which are inherently linked to a normative mode of explanation. In developing this account, I build on the proposal by Levenstein et al. (2023), who advocate for a pragmatic reinterpretation of cognitive neuroscience. According to their framework, mechanistic, normative, and descriptive explanations each have a legitimate place within scientific practice, with the choice among them guided not merely by structural adequacy, but more crucially by their epistemic utility and alignment with specific research questions and aims. Such a conception of neuroscience allows for the integration of diverse knowledge-making practices across multiple explanatory levels, without reducing one kind of explanation to another, and preserving space for the development of future, potentially superior models and accounts.

GoogleMeet link: https://meet.google.com/bhf-bcgj-yto