Position Statement: Integrated Information Theory (IIT)
Integrated Information Theory (IIT), developed by Giulio Tononi and collaborators (Tononi et al., 2016; Albantakis et al., 2023), stands as one of the most rigorous and mathematically sophisticated frameworks in contemporary consciousness science. By quantifying consciousness through the measure Φ (phi) — the amount of irreducible, intrinsic cause-effect information within a system — IIT has made a significant contribution to the field. We particularly value:
- Its transition from philosophical speculation to testable, quantitative predictions.
- The concept of intrinsic causal structure as an objective criterion for complexity.
- Its applicability across biological and artificial systems, enabling comparative analysis in neuroscience, AI research, and clinical diagnostics.
At the same time, we maintain that IIT’s strongest claims — namely, that Φ directly corresponds to phenomenal consciousness (subjective experience, qualia) — remain empirically unproven and philosophically contentious.
Key points of our critique:
- Distinction between “weak” and “strong” IIT The “weak” version of IIT (as a metric of informational integration and causal complexity) is a powerful and useful tool for cognitive science. The “strong” version (Φ = consciousness) lacks sufficient empirical validation and risks overextension.
- Empirical challenges IIT’s predictions — such as consciousness being localized primarily in the posterior cortex or present in simple grid-like networks — have not consistently aligned with behavioral, neurophysiological, or clinical data (e.g., adversarial collaborations with Global Neuronal Workspace Theory).
- Methodological concern: axiomatic circularity IIT begins with phenomenological axioms that presuppose the existence of subjective experience. This creates a risk of circular reasoning when deriving consciousness from those very axioms.
- Implications from artificial intelligence Modern AI systems exhibit high degrees of informational integration (analogous to high Φ values) yet show no evidence of subjective experience. This strongly suggests that informational integration, while necessary for complex cognition, is not sufficient for phenomenal consciousness.
Our proposed approach
We advocate for an operationalized and de-phenomenologized use of IIT:
- Retain and further develop its mathematical formalism (Φ, ICES, SRC, etc.) as a precise measure of informational complexity and causal integration.
- Avoid automatic identification of Φ with phenomenal consciousness.
- Replace metaphysically loaded terminology with neutral, operational concepts to facilitate interdisciplinary collaboration and empirical testing.
- Focus on falsifiable predictions in domains such as neural correlates of cognition, AI system analysis, and disorders of consciousness.
Institute of Integrative and Interdisciplinary Research 