Meeting Them Where They Are At: Using Large Language Models to Lower Barriers to Measuring the Impact of Gender-Based Violence on Mental Health and HIV Outcomes in Girls and Women in Kenya

Mike Baiocchi of Stanford University in the U.S. will use LLMs to analyze conversational interviews with adolescent girls and young women in Kenya to identify causal mechanisms impairing their health within the potential interplay between living with HIV, mental health conditions, and gender-based violence. Working with the Kenyan Medical Research Institute, they will re-identify and enroll a previously studied cohort living with HIV. The new longitudinal dimension of the study will contribute to untangling causes and effects in parallel to the new LLM-based analysis. They will use LLMs to create statistical measures of participants' descriptions of their experiences to help identify the underlying causes, for instance detecting differences between the responses of those experiencing violence or depression compared to those who have not. Such improved understanding will help to design and target appropriate health interventions.