Consciousness Potential simulation code and validation study

Read full text Summarise with VIRAT
Authors: Spencer Howie
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Companion code for:Howie, S. J. (2026b). Quantifying Awareness: A Formal Index for the Physical Substrate of Consciousness. Zenodo. https://doi.org/10.5281/zenodo.19297160 OVERVIEW This deposit is the Python simulation validation study and supplementary analyses for the companion paper. It is self-contained and needs only NumPy and Matplotlib. Three N=12 AR(1) network topologies (Feedforward, Random recurrent, Structured recurrent) are simulated, and the framework's four Consciousness Potential (Phi) components (I_A, 1 minus H_norm, I_M, I_R) are computed on each. The recursive feedback index I_R is operationalized as the normalized spectral radius of the thresholded Granger-causality matrix, the framework's tractable sub-critical estimator for the returning mutual information specification of I_R given in Section IV of the companion paper. The study demonstrates the predicted rank ordering (Recurrent greater than Random greater than Feedforward equals 0) and contrasts the framework's Phi with Tononi-Sporns 2003 integrated information and with literal IIT 3.0. CONTENTS simulation_study_v3.py — Main simulation: computes the four Phi components on the three topologies; writes simulation_figure_v3.png and simulation_results_v3.npy.iit_comparison.py — Tononi-Sporns 2003 integrated information (closed-form Gaussian) on the same three networks; writes iit_comparison_results.json.supplementary_analyses.py — Robustness sweeps, comparison against five simpler metrics, topology generalization (Watts-Strogatz, Barabasi-Albert), component decoupling, and a 30-topology component-correlation matrix; writes supplementary_results.json.supplementary_figures.py — Regenerates the three companion figures.pyphi_iit3_comparison_colab.py — Literal IIT 3.0 (PyPhi) on three 3-node binary networks; designed for Google Colab (see script header); results in pyphi_iit3_results.json.eval_I_R_spec.py — Standalone check of the returning-MI / spectral-radius I_R estimator.run_all.sh — Runs the pipeline end to end.requirements.txt, LICENSE — NumPy + Matplotlib; MIT license. Bundled result files (.npy, .json, .png) reproduce every figure and headline number without re-running anything. REQUIREMENTS Python 3.8 or newer with numpy and matplotlib (both are in standard scientific Python distributions). PyPhi is needed only for the optional literal-IIT-3.0 Colab script. REPRODUCING bash run_all.sh which runs, in order:python3 simulation_study_v3.py (about 3 min)python3 iit_comparison.py (about 30 sec)python3 supplementary_analyses.py --stage all (about 10 min)python3 supplementary_figures.py (about 10 sec) All four scripts honor --out-dir and the PHI_SIM_OUT_DIR environment variable. supplementary_analyses.py accepts --skip-existing to resume an interrupted run. EXPECTED KEY RESULTS Three-topology Phi ratio (Recurrent / Random): 13.4xFeedforward I_R = 0 in every run (rho(W) = 0 for the nilpotent coupling matrix)Phi_TSE: feedforward 1.43, recurrent 0.38, dense 1.70 (TSE-2003 stays positive on feedforward; the framework Phi does not)Literal IIT 3.0 (system-level): near-feedforward 0.0000, recurrent 3-cycle 1.0000, dense XOR 1.8750 VERSION v4 (2026): documentation and packaging hygiene only; no numeric or simulation-behavior change from v3. Removed a cross-paper empirical working note that does not belong in a stand-alone AR-network simulation deposit, and aligned the run commands to python3. Version 3 (May 2026) adopted the spectral-radius I_R as canonical (retaining the reachability proxy as comp_I_R_reach for traceability) and added the robustness, baseline-comparison, TSE, and literal-IIT-3.0 analyses with three new figures. CITATION Software (version-independent concept DOI): 10.5281/zenodo.19226778Companion paper (concept DOI): 10.5281/zenodo.19297160

Fields of Study & Subjects

Philosophy and Theoretical Science Neural dynamics and brain function Functional Brain Connectivity Studies

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
25 March 2026
Language
EN