The Dual Nature of LLM Persona: Aggregated Tendencies and Frame-Dependent Geometry
2026-07-02 • Artificial Intelligence
Artificial IntelligenceMachine Learning
AI summaryⓘ
The authors studied how personality traits shown by large language models (LLMs) like GPT-4o change when questions are asked in different orders or cultural ways. They found that simple average scores for traits are mostly stable even when question order changes, but more detailed patterns of how answers relate to each other are very sensitive to these changes. This means that LLM personalities have two parts: one that stays the same across different 'frames' (question orders or cultures) and one that depends on the frame and shows hidden information. Their work suggests we need to consider these frames when evaluating AI personalities instead of just relying on averages.
LLM personaspsychometric questionnairesBig Five traitswithin-instance correlationSPD manifoldsframe-dependenceIPIP-50personality evaluationgeometric featurestrait stability
Authors
Yuan Yuan
Abstract
Evaluations of LLM personas via psychometric questionnaires typically rely on aggregate scores, discarding within-instance correlation structure. We test whether this geometric structure is intrinsic or frame-dependent. Constructing within-instance correlation matrices from IPIP-50 responses, we analyze geometry on SPD manifolds under manipulated question orderings in GPT-4o simulating American and Chinese-American personas. We find that persona expression comprises two dissociable components: aggregated features (Big Five scores) degrade under randomization (21% drop) but are frame-robust; geometric features (SPD manifold) collapse under frame misalignment (42% drop) but recover substantially (to 84%) under shared frames, surpassing aggregated features (76%). This collapse-recovery pattern reveals that persona geometry is not intrinsic but a frame-dependent coordination pattern encoding information invisible to aggregation. Our findings establish a dual-nature framework for LLM personas, frame-dependent geometry versus frame-robust aggregates, necessitating frame-aware evaluation and challenging static trait conceptions.