Computational and Neuroscientific Objections to Integrated Information Theory

Abstract

Integrated information theory (IIT) equates consciousness to integrated information, a concept that quantifies the interconnectedness of a system’s physical causal properties. However, IIT relies on a metric of complexity for integrated information—the phi (Φ) metric—that is both impractical to compute and insufficient to predict consciousness. Evaluating the Kullback-Leibler divergence for all system bifurcations causes the number of states to grow rapidly, making the calculation of Φ computationally impractical. And because high complexity, non-conscious systems exist (for example, some communication error correction codes have a large Φ but are not conscious) the proposed metric does not reliably indicate consciousness. There are also neuroscientific challenges to IIT. The theory relies on the assumption that all mental states supervene on brain states, implying that consciousness arises solely from physical processes in the brain. We propose that many studies in classical neuroscience point to the immateriality of the intellect and will, challenging the IIT of consciousness. Unlike locomotion, perception, memory, and emotion, higher cognitive powers such as intellect and will lack precise localization in the brain. Intellectual seizures do not happen, and abstract thought cannot be triggered by intraoperative brain stimulation. Intellect remains coherent in split-brain patients, and awareness persists in deep coma and near-death experiences. Studies showing free will operates separately from brain processes also support the immateriality of intellect and will. These observations challenge IIT’s explanation of consciousness. In conclusion, while IIT presents an elegant mathematical framework, it does not fully capture the complexities of human consciousness. Explaining consciousness requires exploring beyond the modeling of material processes in the brain or the integration of information these processes convey.