Level of Consciousness Signatures Across Biological and Artificial Minds: A Unified Framework for Measuring Cognition in Human EEG and Large Language Models

As artificial intelligence systems approach human-level cognitive capabilities, we lack unified frameworks to measure and compare cognition across biological and artificial minds. We introduce the Level of Consciousness (LOC) framework, which defines 13 cognitive func- tions — Thinking, Reasoning, Understanding, Cognition, Emotion, Attention, Sensation, Feelings, Intuition, Energy, Awareness, Mindfulness, and Consciousness — measurable in both human electroencephalography (EEG) and large language model (LLM) hidden states. We validate LOC through four studies: (1) token-level cognitive function detection in 5 LLM architectures (12B–70B parameters), achieving 10–12 of 13 functions statistically sig- nificant per model (p < 0.001) and sentence-level classification at 2.1–3.1× above chance; (2) cognitive task classification in 21 human EEG subjects from the COG-BCI dataset, achiev- ing 2.31× above chance with all subjects significant (p < 0.001); (3) cross-network analysis showing 7 of 13 functions produce same-direction effects in both biological and artificial neu- ral networks; and (4) causal validation via layer isolation, where Consciousness-designated regions alone achieve 6.5× chance detection across 4 model architectures. True Coher- ence scoring reveals that cognitive coherence scales with model size (Llama-70B: 15.4% vs Gemma-12B: 7.4%). These results provide the first empirical evidence that cognitive func- tion signatures are independent of whether the neural network is biological or artificial, with implications for AGI safety monitoring, AI interpretability, and consciousness research.