Authors:	Sims, Andrew
Title:	The problems with prediction : the dark room problem and the scope dispute
Online publication date:	1-Jun-2017
Year of first publication:	2017
Language:	english
Abstract:	There is a disagreement over the scope of explanation for predictive processing. While some proponents think that it is best motivated from—and indeed comprises an explanation of—biological self-organization, others maintain that it should only be a theory of neurocognitive function, or even just of some limited domain of neurocognitive function. Something that these theorists share is an interest in addressing the dark-room problem: at its most naïve, if action is driven by the minimization of surprise then why don’t cognitive creatures act to minimize stimuli in general? The dark-room problem is in fact best conceived as a cluster of related concerns, rather than as a single argument against action-oriented predictive processing. These have to do with: i) whether PP (predictive processing) has any substantive empirical content when it is pitched in very general domains; ii) whether a specification can be given of the environmental niche that action moves the organism to occupy, and which is not the dark room; and iii) whether an adequate account can be given within this specification of exploratory and exploitative behaviours. There are interesting conceptual relations between the dark-room problem and the scope dispute. As the putative scope of predictive processing gets wider (culminating in the free energy principle), the resources that are available for answering the concerns about niche-specification become very rich. But increasingly puzzling problems arise as to the implementation of surprise-minimisation within non-paradigmatically cognitive biological systems. On the other hand, under more restrictive construals of the scope of predictive processing, there are new difficulties standing in the way of niche-specification, and new questions about the interface between surprise-minimisation and model-free cognition undermine the promise of predictive processing as a unifier of theories of neurocognitive function in subordinate domains. In this paper I make explicit the dialectic between proponents and critics in order to show how the two problems are related.
DDC:	100 Philosophie 100 Philosophy
Institution:	Johannes Gutenberg-Universität Mainz
Department:	FB 05 Philosophie und Philologie
Place:	Mainz
ROR:	https://ror.org/023b0x485
DOI:	http://doi.org/10.25358/openscience-646
URN:	urn:nbn:de:hebis:77-publ-566658
Version:	Published version
Publication type:	Buchbeitrag
License:	CC BY-ND
Information on rights of use:	https://creativecommons.org/licenses/by-nd/4.0/
Citation:	Philosophy and predictive processing Metzinger, Thomas
Pages or article number:	380 397
Publisher:	MIND Group
Publisher place:	Frankfurt am Main
Issue date:	2017
Publisher URL:	http://dx.doi.org/10.15502/9783958573246
Publisher DOI:	10.15502/9783958573246
Appears in collections:	JGU-Publikationen