Synthese (2018) 195:2541–2557
S.I. : PREDICTIVE BRAINS
What do predictive coders want?
Received: 6 April 2016 / Accepted: 18 October 2016 / Published online: 25 October 2016
© Springer Science+Business Media Dordrecht 2016
Abstract The so-called “dark room problem” makes vivd the challenges that purely
predictive models face in accounting for motivation. I argue that the problem is a
serious one. Proposals for solving the dark room problem via predictive coding archi-
tectures are either empirically inadequate or computationally intractable. The Free
Energy principle might avoid the problem, but only at the cost of setting itself up as a
highly idealized model, which is then literally false to the world. I draw at least one
optimistic conclusion, however. Real-world, real-time systems may embody motiva-
tional states in a variety of ways consistent with idealized principles like FEP, including
ways that are intuitively embodied and extended. This may allow predictive coding
theorists to reconcile their account with embodied principles, even if it ultimately
undermines loftier ambitions.
Keywords Predictive coding · Free energy principle · Homeostasis · Good regulator
theorem · Extended mind · Explanation
1.1 The dark room problem
Predictive coding (PC) models depict the nervous system as a machine for hierar-
chically minimizing the prediction error between internal models and sensorimotor
input. Such models have successfully captured a variety of speciﬁc sensory and motor
phenomena (Rao and Ballard 1999; Huang and Rao 2011; Clark 2013; Hohwy 2013).
Department of Philosophy, Macquarie University, Sydney, NSW 2109, Australia