Special Clinical Talk - "Developing AI for Psychological Assessment"

Abstract:

Psychologists recognize that multi-method assessment improves the validity of measurement and they strive to incorporate it into research and practice (APA, 2020). Yet assessment continues to overwhelmingly rely on self-report. There are practical reasons for this: Multi-method assessment takes time, effort, and resources. However, “construct validity is the foundation of clinical utility” (Clark & Watson, 2019) and continued reliance on self-report limits the effectiveness of our treatments. It also limits our knowledge of psychological constructs. Finding new ways to implement multi-method assessment into research and practice would be beneficial, and use of AI to assess personality and psychopathology-related behavioral markers (e.g., speech, language, facial features) might serve this purpose. In particular, natural language has been linked to psychological phenomena (Pennebaker et al., 2003). Recent advances in natural language processing (NLP), and specifically the advent of large language models (LLMs), show potential as multi-method assessment tools that might be easily implemented into research and practice and provide automated insights into psychological constructs. Presented here will be description of initial studies in this area and future research directions in NLP and AI.