New Approaches to Characterize Psychological Traits and States

This thesis addresses two central questions about psychological traits (stable individual differences such as personality) and states (transient fluctuations such as momentary emotions). First, how can we quantitatively distinguish traits from states in empirical data? Second, how are traits structured when we infer them about others, and is that structure shared across perceivers from different cultural backgrounds? To address the first question (Chapter 2), I applied statistical models to longitudinal self-report data collected during the COVID-19 pandemic, demonstrating that psychological measures are mixtures of trait-like and state-like components rather than pure categories, and decomposing their variability into temporally stable between-individual differences and within-individual fluctuations. To address the second question (Chapters 3–4), I developed a label-free experimental paradigm to recover the latent structure of trait impressions from faces without relying on experimenter-imposed trait words, revealing a low-dimensional representational geometry that is largely shared across perceiver racial groups. In the discussion (Chapter 5), I situate these findings within a broader measurement framework, comparing self- versus other-inference and examining how traits and states can be studied in self-report, task-based, neurobiological, and experimental-manipulation approaches across humans, nonhuman animals, and large language models. I outline future directions in person–situation interactions, cross-cultural generalizability, and multimodal measurements. Together, these findings provide a quantitative framework for modeling psychological traits and states in both self-evaluation and social perception, advancing a unified view of psychological structure grounded in statistical organization rather than semantic categories.