Integrated information theory (IIT) is characterized as an instance of 'ironic science' that hinders the scientific investigation of consciousness. The theory mistakenly treats a method for measuring network complexity as a full-fledged theory of consciousness, leading to an internal contradiction that drives it toward panpsychism. The author examines the implications of this conceptual error and proposes a path to return the study of consciousness to rational, empirical science.
A new framework using information-theoretic complexity measures, such as integrated information, has been proposed to quantitatively classify states of consciousness, addressing both phenomenological contents and clinical disorders. However, applying these measures to realistic brain networks is difficult due to high computational costs. This article serves as a lookup table of principle-based and empirically tested measures of consciousness, with emphasis on clinical applicability for assisting diagnosis and therapy. It addresses challenges facing these measures with regard to realistic brain networks and suggests possible resolutions.
A complexity-based morphospace with three axes—autonomous, cognitive, and social complexity—can represent both biological and synthetic conscious systems. Awareness corresponds to computational complexity and wakefulness to autonomous complexity. Consciousness is argued to function as an evolutionary game-theoretic strategy, motivating social complexity as a third dimension. The framework yields a taxonomy of four types of consciousness based on embodiment: biological, synthetic, group, and simulated. This classification aids in identifying design principles for engineering conscious machines and in comparing signatures of consciousness across domains relevant to cognitive neuroscience, AI, and biomimetics.
The distributed adaptive control theory of consciousness (DACtoc) proposes that consciousness evolved during the Cambrian period to help agents deal with hidden states of the world, enabling stable multi-agent environments. It functions as an autonomous virtualization memory that serializes and unifies parallel subconscious simulations of hidden states—largely due to other agents and the self—to extract norms, which are then projected as value onto control systems driving action. This functional hypothesis is mapped onto brainstem, midbrain, thalamo-cortical, and cortico-cortical systems. The theory predicts that normative bootstrapping of conscious agents requires an intentionality prior, and suggests human consciousness represents an ultimate evolutionary transition toward autonomy from evolutionary priors.