A Methodology for Using Large Language Models to Map Scientific and Popular Systems for Describing Personality

This article describes the application of large language models for rapidly mapping formal and informal systems for describing personality, using the Big Five and the popular system of socionics as an example. It shows how personality questionnaire data can be used to obtain semantic projections of traits from one system onto another and to evaluate correlations between the trait factors of the two systems. The article describes a method for formalizing personality traits in popular typologies based on self-diagnosis data and machine learning, allowing qualitative typological categories to be translated into quantitative scales. Using data from the SOLTI-160 socionics questionnaire, linear approximations of socionic traits were obtained. Correlation analysis between the Big Five factor projections obtained with the Gemini 2.5 large language model and socionic traits produced results qualitatively consistent with findings from other studies: Big Five Extraversion correlated with socionic extraversion, Agreeableness with ethics and judiciousness, Conscientiousness with rationality, and Openness to Experience with intuition. To assess the reliability of the methodology, the study was repeated using several other large language models, and comparative analysis showed high inter-model agreement in semantic projections. The article also describes a method for qualitatively mapping systems of types and systems of traits within circumplex models. It examines the average location of socionic types in a circumplex model on a plane defined by Agreeableness and Openness to Experience and analyzes the assumption that pairs of "mirror" socionic types correspond to sectors on this plane. The described approach can be used to compare systems in the absence of data from simultaneous diagnosis, to generate hypotheses, and to validate popular personality typologies and integrate them into scientific psychometrics.