Assistant Professor Southern Methodist University Dallas, Texas, United States
Abstract Body Dimensional models of personality disorder are replacing categorical models which have been the standard in the field (Reed et al., 2019). Despite this, research still relies heavily on self-report measures of personality that have inherent limitations such as response bias and limitations of self-insight which prompts the exploration of alternative assessment methods (Paulhus & Vazire, 2007). Advancements in AI offer a unique opportunity to identify objective dimensional markers of personality such as low agreeableness/antagonism. Language is an essential component in understanding the basic psychological and scientific principles underlying human behavior. Natural language processing (NLP) has already been used to predict aspects of personality through linguistic patterns (Park et al., 2015; Simchon et al. 2022). Research indicates that people lower in agreeableness use more negative emotion and swear words (Yarkoni, 2010). However, transformer-based large language modeling (LLM) has not been fully utilized to examine agreeableness/antagonism in clinical interviews. The present study aims to explore the relations between language and personality using novel transformer-based LLMs and topic modeling to understand agreeableness-related language in lengthy (average time = 20 min.) clinical interviews with N = 120 college students and community members oversampled for diversity in race and ethnicity. Language models will be trained on the Big Five Inventory-2 Agreeableness scale and the Elemental Psychopathy Assessment of antagonism. It is expected this study will replicate prior broad associations between agreeableness/antagonism and language and advance it through learning relations between agreeableness/antagonism using more nuanced linguistic features extracted by LLMs and topic modeling. NLP presents a unique opportunity to advance the understanding of personality through more objective, fast, efficient, and cost-effective means. Through NLP, linguistic nuances may provide clinicians with rapid and uniquely valid assessments from which they can improve assessment, monitoring, treatment planning, and ultimately treatment outcomes.
.png)
