Session: Strengthening Social Impact on LGBTQ+ Health Via Artificial Intelligence (AI): A Collaboration between Social Work and Data Science (Society for Social Work and Research 29th Annual Conference)

Artificial intelligence (AI) tools--exemplified by the large language model (LLM) chatbot ChatGPT--have begun to supplant (or become incorporated into) existing digital information search resources such as internet search engines, Wikipedia, and crowdsourcing tools/platforms (e.g., Reddit). The accessibility, user-friendliness, and ostensibly the quality of AI chatbot output has resulted in increased use not only among the general public, but also among social work students, practitioners, educators, and researchers. In principle, social work practice and research could benefit from AI tools since a significant number of social workers are interested in improving health and well-being of lesbian, gay, bisexual, transgender, queer and other sexual/gender expansive individuals and populations (LGBTQ+) as well as the significant representation of LGBTQ+ individuals/communities served by social workers. The field is on the cusp of being transformed by AI, especially given the speed and accelerating nature of advances in AI techniques/tools. But because of this rapidity, the understanding of if/how AI can be incorporated into social work practice, teaching, and research--especially with a marginalized and stigmatized population for which there is a lot of published mis- and disinformation that could feed AI models--is nascent. Furthermore, the understandings, conceptualizations, terminologies, social contexts, and issues among LGBTQ+ communities are continually evolving, arguably as rapidly as AI.

This symposium covers insights, findings, and lessons learned from an ongoing project focused on assessing the feasibility and potential promise of LLM-based AI chatbots to assist current/future social work professionals working with LGBTQ+ populations and in need of science-based knowledge about LGBTQ+ issues and intervention. Paper One is a scoping review of the scientific literature on LLMs as applied or investigated in the context of LGBTQ+ health; thus, it summarizes and characterizes the state-of-the-art on uses, methodologies, and strengths and weaknesses/concerns of LLMs and LGBTQ+ considerations. Paper Two presents a method that rigorously and reproducibly assesses the quality and usability of AI tools specifically for LGBTQ+ issues/populations. Given the rapidity with which AI tools are advancing and disseminated, an emphasis was placed on a method that is "future proof." Thus, the method can not only be used to assess future AI tools/resources, but also other populations and issues, particularly those characterized by vulnerability and oppression. Paper Three provides details on prompt engineering, a critical technique used by the aforementioned assessment method; specifically, it reviews the RISEN (Role, Instruction, Steps, End goal, Narrowing) framework and provides examples in the context of social work and LGBTQ+ issues. Paper Four identifies and articulates not only the need for social workers to participate in the AI space for LGBTQ+ health, but also the value-added--and, particularly for marginalized/minoritized populations, necessary--contribution of social workers to AI researchers and their collaborators. The symposium discussant will also have the opportunity to present and facilitate insights into what skills and training would specifically benefit social workers interested in effectively and ethically using AI and/or contributing to the advancement of AI for social work teaching, practice, and research.