Research Symposium

BIO

I'm Shiv, a 1st year Exercise Physiology major. I'm from Tallahassee and my research interests include AI in healthcare, cancer genomics, and clinical analytics.

Developing a Trustworthy AI Chatbot for Personalized HIV Care

Abstract

People with HIV (PWH) face persistent barriers to optimal care, including challenges with antiretroviral therapy (ART) adherence, retention in care, and engagement with preventive health services. These barriers disproportionately affect minoritized populations and contribute to disparities in viral suppression and overall health outcomes. Digital health interventions, such as conversational agents (CAs) or chatbots, offer a scalable and accessible means to support PWH across the HIV care continuum by providing personalized, real-time health guidance
To address these challenges, we developed a personalized, AI-powered chatbot—Aipaca—designed to deliver real-time, evidence-based HIV care guidance. Leveraging Large Language Models (LLMs), our chatbot integrates clinically validated recommendations to enhance the accessibility, accuracy, and inclusivity of HIV support.
To fine-tune the AI, we curated a comprehensive HIV Knowledge Bank covering the full continuum of HIV care. This included over 1,080 pages of peer-reviewed material, 450 quiz questions, and 1.2 million tokens of synthesized content. We benchmarked both open-source (Llama, Mistral) and closed-source (GPT-4, Gemini) LLMs on clinical accuracy, reasoning, and patient-centered communication. Preliminary results demonstrate the feasibility of using AI chatbots to provide personalized care, including support for mental health, medication safety, and HIV prevention strategies such as PrEP.
Ongoing efforts include expert interviews to refine clinical applicability, pilot testing with PWH, and performance validation to ensure the chatbot’s safety, accuracy, and scalability in real-world healthcare settings. Our findings highlight the potential of AI chatbots to transform HIV care by offering personalized, accessible, and trustworthy support for PWH.

Keywords: Artificial Intelligence, HIV, Chatbot