Introducing a large-scale, data-driven map of social knowledge structures and clinically relevant person profiles

Here, we introduce a large-scale, data-driven framework to systematically characterize social behavior by identifying dimensions of self-reported personality traits and item preferences and deriving latent person-level profiles and linking these patterns to meaningful clinical symptom dimensions. This framework introduces generalizations of item- and person-level social information, which can be used as social knowledge structures across various social decision-making settings. In Study 1, a final sample of N=1,529 participants completed extensive self-report measures of personality traits and item preferences. We used Exploratory Graph Analysis (EGA) to uncover the dimensional structure of these data, identifying four trait dimensions and four preference dimensions. These dimensions were then used in Latent Profile Analysis (LPA) to define trait and preference profiles. The resulting profiles demonstrated external validity, generalizing to independent measures of personality and clinical symptoms. Notably, we identified a “Maladaptive Trait profile” and a “Raw-Produce-Aversion/Texture-Sensitive preference profile,” both associated with elevated autistic traits, greater behavioral rigidity, and lower emotional stability. In Study 2, we assessed the robustness and generalizability of these dimensions and profiles in an independent sample of German-speaking participants (N=649). The dimensional structure showed partial replication across samples. The “Maladaptive Trait profile” demonstrated the strongest cross-sample correspondence, whereas preference profiles were more variable, with broader food preference profiles replicating more consistently than narrower ones. In Study 3, the item dimensions and people profiles were used for social learning. A separate sample of 48 participants completed a social learning task in which they updated beliefs about traits and preferences from representative individuals of identified latent person-level profiles. We found that social learning performance differed between latent profiles. Together, these findings reveal the usefulness of this extensive data-driven approach that links behavioral dimensions and person-centered profiles to individual differences in social cognition and to psychopathology.