Tuesday, October 31 • 10:40am – 11:30am EST

Large language models, powering chatbots such as ChatGPT, Bing, and Bard, hold great promise for individuals with low health literacy. They can be used to simplify and explain complex medical documents, provide medical advice, and even author patient education materials from scratch. However, there are significant risks associated with their use for medical advice, given their proclivity to fabricate false statements. These risks are amplified for those with low health literacy, since these individuals may have a more difficult time composing appropriate prompts, understanding output text that is often at too high a reading level, and recognizing false statements. In this talk, I will discuss these issues and some remediation strategies for leveraging the power of large language models in a safe way to improve the delivery of medical information for individuals across the health literacy spectrum.