Artificial Intelligence and Stigma in Addiction Research: Insights From the HEALing Communities Study Coalition Meetings

Abstract

Objectives: This paper describes how artificial intelligence (AI) was used to analyze meeting minutes from community coalitions participating in the HEALing Communities Study. We examined how often coalitions discussed stigma when selecting evidence-based practices (EBPs), variations in stigma-related discussions across coalitions, how these discussions addressed race, ethnicity, and racial inequity, and whether the frequency of stigma discussions was associated with the proportion of minoritized populations in each community. Methods: We used Natural Language Processing, Machine Learning, and Large Language Models, employing ChatGPT Enterprise to code data, ensuring data security and privacy compliance with the General Data Protection Regulation and HIPAA. Results: Community coalitions varied in the extent to which they discussed stigma during meetings focused on EBPs to reduce overdose deaths. Stigma was mentioned more frequently in the context of medication for opioid use disorder compared with other EBPs. As the percentage of racial/ethnic minority populations increased in a county, so did the strength of the association between discussions of EBPs and stigma. Counties with a greater proportion of racial/ethnic minority populations were more likely to integrate discussions of EBPs with stigma-related issues. Specifically, discussions about stigma were 57% more likely to occur when racial or ethnic disparities were mentioned, compared with when they were not (odds ratio=1.57; 95% CI: 1.22, 2.03). Conclusions: The paper highlights the potential for integrating AI-human collaboration into community-engaged research, particularly in leveraging qualitative data such as meeting minutes. It shows how AI can be used in real-time to enhance community-based research.