Some Conceptual and Empirical Shortcomings of IIT

Abstract

The Integrated Information Theory of consciousness (IIT) has generated much excitement inside and outside the scientific community, and seems to many the leading contender for a satisfactory theory grounded in systems neuroscience. It is a bold theory, one that provides plausible explanations for various recognized neuroscientific facts, makes surprising predictions that go beyond current scientific orthodoxy but are potentially testable, and has inspired development of what appears to be an effective technique for detecting the presence of consciousness in organisms incapable of verbal report, such as non-human animals, neonates, and severely brain-damaged adults. Despite these virtues, IIT appears fundamentally flawed: This paper first revisits some key conceptual and technical issues that have been raised previously but remain unresolved—in particular, issues concerning IIT’s concept of “information” and its approach to the “hard problem”—and then focuses on several empirical phenomena that IIT seems unable to handle satisfactorily. These include: 1. cases of multiple personality or dissociative identity disorder in which complex and overlapping centers of consciousness co-occur in single human organisms; 2. the failure of the intense phenomenology of psychedelic states to be straightforwardly reflected in accompanying neuroelectric activity; and, most critically; 3. the occurrence of profound and personally transformative near-death experiences (NDEs) under extreme physiological conditions such as cardiac arrest, in which IIT predicts that no conscious experience whatsoever should be possible.  These empirical arguments show that IIT itself is untenable, and they apply also to its physicalist competitors. Scientifically and philosophically respectable alternatives, however, are available.