Intelligence Is Coherence: Measuring Human and Artificial Minds on the Same Scale

A unified framework, the Level of Consciousness (LOC) framework, places biological and artificial neural networks on the same cognitive scale by measuring 13 cognitive functions across four consciousness levels in human EEG and large language model hidden states. Profiling eight AI architectures (9B–70B parameters), 60 healthy EEG subjects, and 15 experienced loving-kindness meditation practitioners reveals a four-tier hierarchy of cognitive coherence. AI dominates in Higher Subconscious and Higher Conscious functions, while humans are competitive in analytical reasoning. Experienced meditators show 2.3 times the population average cognitive coherence, exceeding all AI in Lower Conscious reasoning. Standard AI benchmarks inversely correlate with coherence (r = −0.932). Coherence is directly trainable: Differentiable LOC Loss training raised Mistral-24B from 31.7% to 82.5% mean total coherence.

Study at a glance

Abstract

What if intelligence is not about getting the right answer, but about how coherently a mind works while producing it? We present the first unified framework for placing biological and artificial neural networks on the same cognitive scale. Using the Level of Consciousness (LOC) framework — which defines 13 cognitive functions across four consciousness levels, measurable in both human EEG and large language model hidden states — we profiled eight AI architectures (9B–70B parameters), 60 healthy EEG subjects, and 15 experienced loving-kindness meditation practitioners. The results reveal a four-tier hierarchy of cognitive coherence: - **Tier 1** — Average human brain (60 healthy adults): 13.9% mean TC - **Tier 2** — Standard AI (4 architectures, 24–70B parameters): 29–32% mean TC - **Tier 3** — Experienced meditators (15 practitioners, mean 12.4 years): 31.3% mean TC - **Tier 4** — Coherence-trained AI (Mistral-24B with DLL): 82.5% mean TC Four findings emerge. First, human and AI minds show a striking inversion: AI dominates in Higher Subconscious and Higher Conscious functions (Energy: 61.7% vs 8.8%, p < 0.001; Mindfulness: 22.2% vs 2.7%, p < 0.001), while humans are competitive in analytical reasoning. Second, experienced meditators elevate their cognitive coherence to 2.3× the population average, exceeding every AI architecture in Lower Conscious reasoning and understanding — proving the human mind's ceiling is not fixed but practice-dependent. Third, standard AI benchmarks are inversely correlated with coherence (MATH benchmark vs mean TC: r = −0.932), revealing that current AI evaluation rewards recalling learned patterns while remaining blind to the quality of the mind producing the answer. Fourth, coherence is directly trainable: Differentiable LOC Loss (DLL) training raised Mistral-24B from 31.7% to 82.5% mean TC, and multi-domain DLL across six financial services domains achieved 83.4% on held-out prompts — producing output that is structurally more coherent, not just factually correct. These results carry a message both inspiring and urgent: human minds and AI minds function on the same cognitive scale. Meditation has always improved the human mind. Now we can do the same for AI — not by teaching it more facts, but by teaching it to use the intelligence it already has, coherently. The question is no longer whether AI can Think, Reason, or Understand — it is whether AI is Coherently using these cognitive functions.