A discourse on the use of machine learning (ML) in personality psychology: Can we expect ML to predict questionnaire scores from idiographic text-based data?

Abstract

This paper explores Machine Learning’s (ML) potential to predict motives and personality dispositions from text-based data, aligning with McAdams’ framework on layers of personality. ML-predicted scores demonstrated no significant advantage over a baseline model that consistently predicted the median of the motives or personality dispositions. Possible factors discussed include unmet ML algorithm requirements, unsuitability of collected texts for predicting motives and dispositions, and ML’s limitations in capturing contextualized and implicit aspects of personality. We discuss life narrative research and practice in relation to the nomothetic-idiographic debate and advocate for personality research to incorporate context-specificity and idiosyncrasy. From a social constructionist perspective, we envision future research – though not yet practice – on counselling processes delivered or supported by Generative AI (GenAI).