AI models predict human awareness in zero gravity using EEG and LLMs

Earth’s gravity is so fundamental that our brains develop an internal model to anticipate its effects on perception and action. A new paper from researchers Bakytzhan Alibekov, Alina Gutoreva, and Elisa Raffaella-Ferre introduces a computational framework that quantifies how this model shifts under different gravitational loads. The framework, trained on open-access parabolic flight data, splits into two components: CorticalG, a lightweight multilayer perceptron that predicts changes in EEG frequency bands, and PhysioG, which uses independent Gaussian process models to capture heart rate variability, electrodermal activity, and motor control responses.

To bridge the gap between raw physiological data and subjective experience, the team used Claude 3.5 Sonnet to simulate narratives describing alertness, bodily awareness, and cognitive state across zero gravity, partial gravity (Moon and Mars), and hypergravity. The combined approach offers a predictive tool for assessing human performance and resilience in spaceflight. Published on arXiv (v2, July 2026), the paper is 60 pages with 5 figures and two datasets, and sits at the intersection of neuroscience, machine learning, and signal processing. It could inform the design of future crewed missions by anticipating how astronauts will feel and function in alien gravity environments.