Beyond the brain: a computational MRI-derived neurophysiological framework for robotic conscious capacity.

A new metric called the Attribution Consciousness Index (ACI) estimates the likelihood that neural activity supports conscious processing by balancing measures of dynamic information and complexity. Using brain simulations and artificial neural networks, the ACI follows a log-normal distribution, enabling robust thresholding: values above 10 correspond to over 90% probability of conscious emergence. The framework also applies to artificial systems, explaining 38.4% of variance between biological and AI-derived patterns. While not measuring subjective experience, the ACI predicts when neural or artificial conditions are poised to sustain consciousness, with potential applications in disorders of consciousness, anesthesia monitoring, neurorehabilitation, and evaluating neuroprosthetics, generative AI, and robotics.

Study at a glance

Abstract

Explaining when neural activity supports conscious processing remains an unresolved question in neuroscience. Current frameworks describe correlates of consciousness but rarely provide thresholds to predict its emergence or recovery. We introduce the Attribution Consciousness Index (ACI), a metric that estimates the generative potential of consciousness by balancing measures of dynamic information (Φ) and complexity (κ) expressed as a normalized odds ratio. Using the empirically validated Connectome-76 within The Virtual Brain, we ran 500 resting-state simulations, selecting lowest-entropy regions to capture informative subnetworks. The ACI followed a log-normal distribution and highlighted hubs-cingulate cortex, dorsomedial prefrontal cortex, hippocampus, and amygdala-implicated in conscious processing. To test generality, we extended the framework to an artificial neural architecture with hierarchical modules, nonlinear Hebbian plasticity, and controlled entropy. Across 1921 executions, the ACI conformed to log-normal laws, enabling robust thresholding. Kernel ridge regression showed predictive validity: AI-derived ACI patterns explained 38.4 % of variance in human ACI distributions, revealing transferable principles between biological and artificial circuits. This extension indicates that ACI can guide artificial-consciousness models implementable in robotics, providing measurable criteria for when robotic systems might sustain conscious-like states. Two contributions are novel. First, ACI thresholds provide interpretable decision points: values above 10 correspond to probabilities greater than 90 % for conscious emergence. Second, the framework offers translational applications-from prognosis in disorders of consciousness, anesthesia monitoring, and neurorehabilitation to evaluating neuroprosthetics, generative AI, and robotics with conscious capacities. While ACI does not measure subjective experience, it predicts when neural or artificial conditions are poised to sustain it.