Architectural Variance Dominates Stimulus Variance in Six of Seven Substrate-Agnostic Consciousness Operationalizations
Open MIND May 28, 2026 Nate Travis
Seven mathematical theories of consciousness (Integrated Information Theory, Global Workspace Theory, Attention Schema Theory, Higher-Order Thought, Predictive Processing Theory, Quantum Information Theory, and the Free Energy Principle) claim substrate-independence—their formulas should work on any system, whether biological or artificial. To test this, researchers applied substrate-agnostic versions of each theory to activity patterns from AI language models and human brain scans processing the same narrative stories. At the population level, AI and human scores overlapped for several theories, but per-stimulus correlations across substrates were near zero for six of seven theories. Only Global Workspace Theory showed a modest correlation (r = +0.365), though it varied across AI architectures. The findings suggest that current operationalizations cannot determine whether the AI substrates carry stimulus-relevant information invisible to these measures or lack it entirely.