Information Integration Theory and The Question of Consciousness: Can Artificial Intelligence Become Self-Aware?

Abstract

While complex information processing has produced remarkable artificial intelligence (AI) systems, one of the central fears among researchers and society at large is whether AI could ever become conscious or self-aware, and if so, whether it might act independently of human command. This concern has fuelled intense debate about whether increasing complexity in information processing, as the information integration theory (IIT) proposes, could eventually give rise to consciousness in AI systems. In this study, we probe whether IIT claims can, in principle, create consciousness, and if so, under what conditions; and if not, why not. We review the strongest versions of the complexity hypothesis within IIT. Our findings suggest that while complex information processing is necessary for intelligence, it is not sufficient for consciousness. From the substance dualist perspective adopted in this study, AI systems may simulate awareness, but there is currently no empirical evidence that they possess genuine subjective experience. We argue that information, whether conceived as signal or semantics, is not self-interpreting and cannot by itself account for subjectivity. Within this philosophical framework, the brain may be the instrument of awareness rather than its ultimate origin. We conclude that, based on the substance dualist perspective adopted in this study and current empirical evidence, scaling AI models appears to increase competence and control, but not establish genuine self-awareness. Therefore, information science should treat AI as powerful but non-conscious systems, focusing on governance, transparency, and ethical responsibility rather than assumed sentience.