10月14日にモナシュ大学 土谷尚嗣先生による「The Structure of Integrated Information Correlates with the Contents of Consciousness」と題した講演を開催いたします。皆様のお越しをお待ちしています。
日時 10月14日 10時30分-12時00分
場所 CiNet 1F 大会議室
The Structure of Integrated Information Correlates with the Contents of Consciousness
School of Psychological Sciences
Faculty of Medicine
Nursing and Health Sciences
The Integrated information theory (IIT) aims to explain how our conscious phenomenology is supported by the mechanisms in the brain. The IIT predicts that a pattern of integrated information generated by sub-mechanisms of the system corresponds to the contents of consciousness. Previously, we invented phi*, a well-behaving approximation of the originally proposed theoretical quantity, phi. Here, we set out to test the prediction by measuring a pattern or structure of phi*, based on field potentials recorded with intracranial electrodes in awake epilepsy patients. The patients performed several visual psychophysical tasks, including backward masking and continuous flash suppression. We found a structure of phi*, constructed from the face selective areas in the ventral or lateral temporal lobe, without any supervised learning, classified subjective visual experience of faces, rather than properties of physical visual input of faces. We ruled out a possibility that high dimensionality of the phi*-pattern was sufficient for such categorization by analysing high-dimensional pattern derived by other quantities (e.g., entropy and mutual information) as well as phi* structure derived from other areas that are not selective to faces; both cases poorly reflected subjective experience compared to phi* structure derived from the face selective areas. In sum, without any training or decoding, high-dimensional phi* structure naturally mirrored subjective experience of faces, supporting an idea that IIT as a possible solution to the Hard problem; why phenomenology has to arise from particular mechanisms in certain states.