Gravity-Awareness: Deep Learning Models and LLM Simulation of Human Awareness in Altered Gravity

Earth’s gravity has fundamentally shaped human development by guiding the brain’s integration of vestibular, visual, and proprioceptive inputs into an internal model of gravity: a dynamic neural representation enabling prediction and interpretation of gravitational forces. This work presents a dual computational framework to quantitatively model these adaptations. The first component is a lightweight Multi-Layer Perceptron (MLP) that predicts g-load-dependent changes in key electroencephalographic (EEG) frequency bands, representing the brain’s cortical state. The second component utilizes a suite of independent Gaussian Processes (GPs) to model the body’s broader physiological state, including Heart Rate Variability (HRV), Electrodermal Activity (EDA), and motor behavior. Both models were trained on data derived from a comprehensive review of parabolic flight literature, using published findings as anchor points to construct robust, continuous functions. To complement this quantitative analysis, we simulated subjective human experience under different gravitational loads, ranging from microgravity (0g) and partial gravity (Moon 0.17g, Mars 0.38g) to hypergravity associated with spacecraft launch and re-entry (1.8g), using a large language model (Claude 3.5 Sonnet). The model was prompted with physiological parameters to generate introspective narratives of alertness and self-awareness, which closely aligned with the quantitative findings from both the EEG and physiological models. This combined framework integrates quantitative physiological modeling with generative cognitive simulation, offering a novel approach to understanding and predicting human performance in altered gravity

💡 Research Summary

This paper introduces a two‑stage computational framework designed to capture how humans adapt to altered gravity both physiologically and subjectively. The first stage is a lightweight multi‑layer perceptron (MLP) that predicts changes in the power of major EEG frequency bands (alpha, beta, theta) as a function of gravitational load (g‑force), flight phase (ascent, steady‑state, descent), and subject demographics. Training data are derived from published parabolic‑flight studies; reported mean EEG values serve as anchor points, and Gaussian noise is added to generate a continuous training set. The MLP consists of three hidden layers (64‑32‑16 neurons) with ReLU activations, L2 regularization, and the Adam optimizer. Ten‑fold cross‑validation yields a mean absolute error below 0.07 µV², outperforming linear baselines by roughly 30 %. Notably, the model predicts a marked reduction in alpha and beta power and a modest increase in theta power at hyper‑gravity (1.8 g), reflecting heightened arousal and altered cognitive load.

The second stage comprises a suite of independent Gaussian Processes (GPs) that model broader physiological variables: heart‑rate variability (HRV) metrics (SDNN, RMSSD, LF/HF ratio), electrodermal activity (EDA) measures (mean conductance, peak frequency), and motor behavior derived from accelerometer‑based gait analysis. Each GP uses an RBF kernel with hyper‑parameters optimized via maximum likelihood estimation. Data augmentation follows a Bayesian sampling approach based on literature‑reported means and standard deviations. The GP ensemble reproduces a non‑linear, U‑shaped EDA response (minimum at 0 g, maximum at 1.8 g) and a decreasing HRV trend with increasing g‑load, while gait cadence shortens under hyper‑gravity and lengthens in micro‑gravity.

To bridge quantitative predictions with subjective experience, the authors employ Claude 3.5 Sonnet, a large language model (LLM), to generate introspective narratives. For each gravity condition, the physiological outputs of the MLP and GPs (e.g., alpha power = 12.3 µV², RMSSD = 42 ms, EDA = 0.15 µS) are fed into a prompt that asks the model to describe the person’s level of alertness and perceived effects of the gravity change. The LLM produces coherent, human‑like accounts: “In 0 g I feel a floating sensation mixed with a slight unease,” “At 0.38 g (Mars) the weak pull gives my body a new feedback loop, making me feel more focused,” and “At 1.8 g the body is compressed, arousal spikes, and I become hyper‑aware of surrounding sounds.” These narratives align closely with the physiological trends predicted by the MLP and GP models, demonstrating that the LLM can translate objective biometric states into plausible subjective reports.

Key insights include: (1) EEG band power varies non‑linearly with g‑load, with alpha/beta suppression indicating increased arousal under higher gravity; (2) HRV and EDA provide complementary autonomic markers that the GP framework captures despite sparse data; (3) LLM‑generated narratives faithfully reflect the underlying biometric profile, offering a novel method to simulate human consciousness in altered environments; and (4) the integrated framework can serve as a predictive tool for astronaut performance, low‑gravity habitat design, and high‑g training protocols. The authors suggest future work involving real‑time data collection during flight, refinement of prompt engineering, and extension of the model to include cognitive task performance metrics.