Beyond IIT: (how) can we model the evolution of consciousness?

Abstract

Tononi et al.'s "integrated information theory" (IIT) postulates rules for assigning measures Phi and qualia types Q of consciousness to classical information networks. We consider whether IIT is compatible with Darwinian evolution. We argue that an IIT-like theory that assigns consciousness to physical systems by relatively simple mathematical rules poses extraordinary ?ne-tuning problems.For example, why, among all possible lawlike theories of consciousness, do we have one that makes us conscious of a high-level narrative of our environment and actions, so accurate that it appears to us to cause our behaviour?We introduce IIT+, a class of extensions of IIT in which Phi and/or Q influence the network dynamics. We argue that IIT+-like theories, unlike IIT-like theories, offer at least partial explanations of how some key features of consciousness evolved. We conclude that if one takes seriously Darwinian evolution and the case for an IIT-like theory, one has to take seriously the case for an IIT+-like theory.