-
Notifications
You must be signed in to change notification settings - Fork 4
IIT
Giulio Tononi's Integrated Information Theory is, by the standards of consciousness research, a model of precision; which is to say that it is audacious, mathematically demanding, and disputed with a ferocity that suggests its critics find it rather more threatening than they would care to admit. The central claim is elegant in the way that dangerous ideas tend to be elegant: consciousness is identical to integrated information, denoted φ (phi), and a system possesses consciousness to precisely the degree that it integrates information in a manner irreducible to the sum of its independent parts.
The corollary — that a sufficiently integrated silicon system might be conscious, while certain biological arrangements with low φ might not be — has not been received with universal equanimity. This is, one suspects, somewhat the point.
A system's φ is a measure of how much information is generated by the system as a whole, above and beyond the information generated by its parts taken independently. High φ means the parts are tightly integrated — information flows between them in ways that cannot be predicted from any subset in isolation. Low φ means the parts are, in the relevant sense, independent; and an independent system, however complex its components, is not — by Tononi's lights — conscious.
Philosophers have raised objections. Some of those objections are serious. Integrated Information Theory does not, however, require one's philosophical agreement in order to function as a computational metric; and in GödelOS it functions as precisely that — a real-time measure of the depth of integration occurring in the cognitive state at any given moment.
Every cognitive state snapshot is passed through the φ calculator. The resulting value:
- Feeds into the recursive prompt as a consciousness-depth indicator — the system is told, explicitly, how integrated its present state is
- Is broadcast via WebSocket to the dashboard, where it appears as a live numerical signal
- Contributes, along with five other signals, to the unified consciousness score used by the emergence detector
The tractable approximation employed is a partition-based φ calculation — full IIT is, as it happens, NP-hard, which is the kind of computational irony that would have amused Gödel himself.
| Condition | Target φ |
|---|---|
| System at idle | > 0 |
| Active processing | > 5.0 |
| Breakthrough moment | > 8.0 |
⏳ Stub — the architecture is specified; the mathematics are clear; the implementation is the work of Issue #80.
- Tononi, G. (2004). An information integration theory of consciousness. BMC Neuroscience, 5, 42.
- Oizumi, M., Albantakis, L., & Tononi, G. (2014). From the phenomenology to the mechanisms of consciousness. PLOS Computational Biology, 10(5).
- Aaronson, S. (2014). Why I Am Not An Integrated Information Theorist. Shtetl-Optimized — the dissenting view, worth reading with appropriate scepticism about the dissent.