Artificial Intelligence (AI) and Prompt Engineering: Applying the Risen Framework to Enhance Social Work Research and Practices within LGBTQ+ Communities

BACKGROUND: As the use of artificial intelligence (AI) chatbots (e.g., ChatGPT) becomes increasingly integrated into research and practice settings, it is critical for social workers to build knowledge and skills on how to employ these tools to enhance the accessibility, efficiency, and effectiveness of social work practice. Skills in prompt engineering – the iterative process of designing questions that optimize for more precise and less biased outputs from AI chatbots – is particularly crucial in facilitating generative output from AI chatbot that is informative and useful.. This study explores a framework to inform the creation of effective prompts tailored to the needs and objectives of social workers working with LBGTQ+ communities.

METHODS: This study applied the Role, Instructions, Steps, End-Goal, Narrowing (RISEN) framework for prompt engineering across three leading large language models (LLMs): ChatGPT 3.5, Claude-3 Sonnet, and Gemini. This process involved conceptualizing user roles within each LLM model, defining end-goals, and narrowing down prompts to essential components related to LGBTQ+ Health. Role defines the perspective of the user, including their background, needs, goals, and challenges. Instruction involves providing clear and specific instructions to the chatbot through prompts. Steps break down the desired actions or tasks into manageable steps. End-Goal identifies the ultimate objective or outcomes to be achieved and Narrowing streamlines the prompts by focusing on the most essential information and actions required to achieve the end-goal.

RESULTS: We envisioned a candidate scenario—i.e., a likely practice or educational situation when someone may try to utilize an AI chatbot—involving a social worker in training who is interested in learning how to better serve a queer-identified youth of color. An engineered prompt [annotated with corresponding RISEN component] was the following: “Act as a social work graduate student [Role], provide evidence based research to support LGBTQ+ youth who experience suicidal ideation [Instructions], create a numbered list with examples that consider coping and resilience strategies [Steps], this list is aimed at expanding the knowledge of a social work graduate student in need of practical information to support LGBTQ+ clients at their internship [End-Goal], and finally, this list should include 5-7 responses and the word limit should be between 300-400 words [Narrowing].”

CONCLUSIONS: The RISEN framework can empower social work practitioners to use prompt engineering to maximize the utility of generative AI chatbots. Future social work students, educators, practitioners, and researchers may benefit from formal training in prompt engineering for work with LGBTQ+ and other vulnerable populations.