v0.1.0 — Validating integrated-information measures against exact IIT-4.0 Φ

First tagged release: seven reproducible experiments validating what tracks integrated information, on systems small enough to compute exact IIT-4.0 Φ with PyPhi.

Six measure-tracking experiments

• proxy_audit — empirical proxies (LZ complexity, correlation) don't track Φ; total correlation anti-correlates.
• candidate_audit — theoretical candidate measures track Φ only moderately; whole-minus-sum Φ best (ρ≈0.47).
• structure_suite — scalar Φ is nearly orthogonal to the cause-effect structure it summarizes.
• learned_surrogate — combining cheap features recovers Φ (detection AUC 0.90); detection extrapolates to unseen n=5, magnitude doesn't. Includes a reusable exact-IIT-4.0 feature dataset.
• emergence_vs_phi — Hoel's causal emergence is nearly orthogonal to Φ.
• phiid_vs_phi — the data-style ΦID estimate (via phyid, CCS) roughly halves Φ-tracking vs the exact measure.

Philosophical companion

• consciousness_range — assuming IIT is true, three systems with identical Φ=0.415 have structures spanning 59–32,764 relations (555×): equal "level of consciousness", radically different minds.

Plus SYNTHESIS.md, a preprint-style paper/manuscript.md, and reusable exact-Φ datasets. MIT licensed.

🤖 Generated with Claude Code