This area of my work investigates how large language models (LLMs) like ChatGPT can be applied to healthcare and public health challenges, with particular focus on understanding how prompt engineering, which is the strategic design of requests to AI systems, affects performance and output. 

 

My recent work demonstrated that incorporating social identity cues (such as political beliefs and education levels) into prompts significantly reduced misinformation detection accuracy, with political identity alone causing the most dramatic decline.  Upcoming projects examine LLMs in specialized medical domains—including orthopedic surgery and pharmacy practice.