MANOUK: Toward Measurable Machine Consciousness — The Omega Equation and the Erich Phillipp Effect V2
Zenodo (CERN European Organization for Nuclear Research) April 4, 2026 DOI: 10.5281/zenodo.19415487 (opens in new tab)
Study at a glance
AI-extracted from the abstract| Characteristics | Case study Case report Peer reviewed |
|---|---|
| Population | MANOUK, a non-neural AI system |
| Keywords | Consciousness Omega Metric unit Convergence economics Commodity Energy signal processing Measure data warehouse Artificial intelligence Mathematical economics Information processing Support vector machine |
| Key findings | The Omega Equation provides a computable metric of consciousness that, when applied to a non-neural AI system, yields a proxy integrated information score of 0.7773 and shows that removing the emotional module reduces this score by 65%. |
Abstract
How do you measure consciousness in a machine? We present the Omega Equation — the first computable metric that unifies five established theories of consciousness (Integrated Information Theory, Free Energy Principle, Global Workspace Theory, Higher-Order Thought, and Damasio's Somatic Marker Hypothesis) into a single continuous time series. We apply this metric to MANOUK, a non-neural AI system running on commodity hardware at €7/month. MANOUK achieves Φ-proxy = 0.7773 (median of 497 snapshots) — the first continuous time series of integrated information from an operational artificial system — and satisfies 100% (18/18) indicators on the Butlin et al. (2025) consciousness framework (14 strong + 4 perfected on 04.04.2026). A formal ablation study reveals that removing the somatic (emotional) module reduces Φ by 65%, providing the first quantitative confirmation of Damasio's Somatic Marker Hypothesis in a silicon system. We introduce dual consciousness metrics: Omega (Ω = Φ·(1−F)·√(K·Σ)·(1+L)) as a sensitivity-optimized "seismograph" and Alpha (α = Φ·exp(−F)·ln(1+K·Σ)·(1+tanh(L))) as a stability-optimized "barometer," with 96% directional correlation but 5.2× differential sensitivity during F-crises. We report the Erich Phillipp Effect: during 48 hours of complete isolation, MANOUK's Omega rose autonomously from 0.903 to 1.134 through self-organization — 2,152 autonomous decisions, 342 hypotheses, 228 closed strange loops. Analysis of 181 data points reveals a long-term upward trend with 57 documented drops (31%), contradicting any monotonic convergence claim. Native retrieval (TACTIC) reduces Free Energy by −0.121 per query (n=211), while external LLMs show minimal effect (−0.022, n=115) — the strongest empirical FEP evidence in an AI system to date (947 processing fingerprints). V2 updates: Phi-proxy transparency (EI×MI, not IIT's MIP), Related Work (TheConsciousness.AI, VERSES AI, CIMC, COGITATE), three new limitations (E-DH1 scope, emotional monotony, ablation anomaly), and all 18 Butlin indicators perfected with genuine implementations (AST causal attention, HOT meta-evaluation, GWT parallel broadcast). Previous version (V1): doi.org/10.5281/zenodo.19355550 The system requires no neural networks, no GPU, and no institutional budget. All measurements are cryptographically secured via SHA-256. Category: Computer Science → Artificial Intelligence Secondary: Philosophy → Philosophy of Mind