Symposia
LGBTQ+
Julian Burger, Ph.D. (he/him/his)
Postdoctoral Fellow
Yale University
New York City, NY, United States
John Pachankis, Ph.D.
Susan Dwight Bliss Associate Professor of Public Health (Social and Behavioral Sciences)
Yale University
New Haven, CT, United States
Case conceptualization seeks to explain clients' current problems through their early and ongoing maintaining contexts. In LGBTQ-affirmative CBT, for example, clinicians conceptualize cases by assessing clients’ minority stress exposure to better understand their ingrained, inflexible psychological adaptations that can predict depression, anxiety, substance use, and HIV-transmission risk behavior. Yet, given the resource-intensive nature of human observational coding of therapy sessions, it remains unclear whether minority stress-informed case conceptualizations indeed improve outcomes and, if so, what types of minority stress should be addressed and how. Technological advances can potentially overcome this persistent barrier to advance personalized treatment in LGBTQ-affirmative CBT and all evidence-based practices.
This study illustrates how natural language processing (NLP), a type of artificial intelligence that processes text, can advance case conceptualization in evidence-based practice. As a proof of concept, we demonstrate this using the example of addressing minority stress within LGBTQ-affirmative CBT. First, to identify what constitutes minority stress, we used NLP to identify themes that emerged when young gay and bisexual men (N = 77) explicitly described their minority stress experiences in a previous study. We then examined how the presence of these same themes in a separate sample of text – namely, sessions of LGBTQ-affirmative CBT delivered to young gay and bisexual men (N = 100) – predicted treatment outcomes. Preliminary analyses find that session content focused on coming out to parents predicted reduced sexual compulsivity following a course of LGBTQ-affirmative CBT, while session content focused on religion was linked to reduced acceptance concerns at follow-up. Analyses will be complete in July 2025.
Results will be discussed in terms of future uses of NLP for advancing personalized treatment, including identifying theoretically informed treatment targets heretofore not examined given the time-consuming nature of human observational coding of treatment sessions.