Psychometrics with AI Foundation Models

Psychometrics is a subfield of psychology concerned with the objective measurement of latent constructs, such as values, personality, mental disorders, and intelligence. AI Foundation models (FMs)-notably large language and multimodal models-are reshaping this field by capturing rich linguistic and behavioural regularities at scale. This paper reviews the emerging synergy between FMs and psychometrics. First, we map practical applications of FMs across the measurement pipeline. Next, we describe key methodologies for enhancing FM performance in psychometric contexts via prompt engineering, fine-tuning, and agentic frameworks. We then examine the theoretical implications of FMs for psychometrics. Finally, we chart risks and offer actionable recommendations for the effective, rigorous, and ethical implementation of FMs in psychometric research and practice. Realizing the potential of FMs in psychometrics demands ongoing interdisciplinary collaboration and continued alignment between advances in computational and measurement sciences.