Human beings and artificial intelligence models learn in fundamentally different ways. Humans update beliefs through repeated exposure, a Hebbian process that creates cognitive “inertia” causing beliefs to resist change. Large language models (LLMs), by contrast, do not update parameters at inference and are unburdened by cumulative experience. Despite growing interest in comparing human and machine cognition, no prior study has placed both on a shared measurement scale permitting direct quantitative comparison. We address this gap using the Galileo Method, a multidimensional scaling technique that preserves ratio-scaled distances in their native metric and which admits non-Euclidean geometries, applying it to dissimilarity judgments from both human respondents and three LLM systems (Claude, DeepSeek, and ChatGPT-5) across identical concept sets. Human respondents display inertial pseudo-Riemannian signatures consistent with Hebbian learning, while LLMs exhibit what we term “massless” reasoning dynamics, repositioning concepts fluidly without the friction of prior reinforcement. This pattern is consistent across all three systems, suggesting it reflects architectural properties of transformer-based models rather than implementation-specific idiosyncrasies. This study provides the first direct human-LLM comparison on a shared, geometrically unconstrained measurement scale, demonstrates detectable architectural signatures across Hebbian and LLM cognition, and establishes a methodological template for the emerging field of Comparative Cognitive Architecture.
Iacobucci et al. (Fri,) studied this question.