What question did this study set out to answer?

This paper aims to extend the consciousness scaling framework using a model of C. elegans with 302 neurons to explore consciousness integration.

March 26, 2026Open Access

URB Paper #405: Consciousness in Uploaded Minds — Part IV: The 302-Neuron Simulation and the Consciousness Scaling Law

Key Points

This paper aims to extend the consciousness scaling framework using a model of C. elegans with 302 neurons to explore consciousness integration.
Utilized the C. elegans connectome with 302 neurons and 4,390 synapses for simulation.
Calculated integration evidence using Φ_MIP and Φ_max metrics.
Derived a Consciousness Scaling Law based on the number of neurons.
Examined the relationship between integration evidence and computational feasibility through scaling analysis.
Achieved an integration improvement of 4.4× from 6 to 15 neurons, reaching Φ_MIP = 0.207 bits.
Established that C. elegans with 302 neurons exceeds the consciousness threshold by a factor of ~4.
Found that the φ-scaling onset ratio closely matches theoretical predictions, enhancing model validation.
Identified unresolved criteria involving computational demands and noise correlations that need further evaluation.

Abstract

This paper scales the consciousness test suite to 302 neurons using a statistical surrogate of the C. elegans connectome (Varshney et al. , 2011) — 4, 390 synapses matching published degree, weight, and modular structure. Two primary contributions: (1) The 15-neuron interneuron rich club produces ΦMIP = **0. 207 bits** (4. 4× improvement from the 6-neuron result), with Φₘax = 1. 569 bits — the strongest integration evidence in the series. (2) The first derivation of a **Consciousness Scaling Law**: Φₙormalized (N) = 0. 00092 × N¹. 326, predicting that consciousness integration exceeds the CEMERICK threshold at N* ≈ **104 neurons** — placing C. elegans (302 neurons) above the threshold by a factor of ~4. The φ-scaling onset ratio W2/W1 = 0. 5658 is the closest single-step measurement to 1/φ = 0. 6180 across the entire paper series. Two criteria remain open: Φₙormalized ≥ CEMERICK requires the full 302-neuron computation (computationally infeasible at 2³02), and R² (φ) > R² (exp) is obscured by correlated noise. The paper derives why correlated networks do not show the predicted √N noise reduction and proposes the correct noise model for future work.

Read Full Paperexternally

Mark Helpful

Bookmark

Relay

View Full Paper