Search papers, labs, and topics across Lattice.
This paper investigates speech embeddings from clinical interviews of psychotic patients and healthy controls using Lyapunov exponent (LE) spectra to model language production as a dynamical process. The study computes LE spectra from word-level and answer-level embeddings generated by two large language models, finding that word-level embeddings show contracting dynamics while answer-level embeddings exhibit positive LEs and higher-dimensional attractors. The LE spectra effectively differentiate psychotic from healthy speech, suggesting that nonlinear dynamical invariants of speech embeddings can serve as a stable probe of disordered cognition.
Lyapunov analysis reveals that speech embeddings from psychotic patients exhibit distinct dynamical signatures compared to healthy controls, offering a novel, physics-inspired approach to understanding disordered cognition.
We analyze speech embeddings from structured clinical interviews of psychotic patients and healthy controls by treating language production as a high-dimensional dynamical process. Lyapunov exponent (LE) spectra are computed from word-level and answer-level embeddings generated by two distinct large language models, allowing us to assess the stability of the conclusions with respect to different embedding presentations. Word-level embeddings exhibit uniformly contracting dynamics with no positive LE, while answer-level embeddings, in spite of the overall contraction, display a number of positive LEs and higher-dimensional attractors. The resulting LE spectra robustly separate psychotic from healthy speech, while differentiation within the psychotic group is not statistically significant overall, despite a tendency of the most severe cases to occupy distinct dynamical regimes. These findings indicate that nonlinear dynamical invariants of speech embeddings provide a physics-inspired probe of disordered cognition whose conclusions remain stable across embedding models.
