LAUSR

UNSTRUCTURED With the launch of Open AI's GPT-3 model in June 2020 came a new era for conversational chatbots (Open AI, n.d.). While there are chatbots who do not use AI, conversational chatbots integrate AI language models that allow for back-and-forth conversation between an AI system and a human user. GPT-3, since upgraded to GPT-4, harnesses a natural language processing (NLP) technique called sentence embedding, and allows for conversations with users that are more nuanced and realistic than ever before. The launch of this model came in the first few months of the COVID-19 pandemic, where increases in healthcare needs globally combined with social distancing measures made virtual medicine more relevant than ever. GPT-3 and other conversational models have been used for a wide variety of medical purposes, from providing basic COVID-19 - related guidelines to personalized medical advice and even prescriptions. The line between medical professionals and conversational chatbots somewhat blurred, notably in hard-to reach communities where the chatbot replaced face-to-face healthcare (Chagas et al, 2021). Considering these blurred lines, and the circumstances accelerating the adoption of conversational chatbots globally, we analyze the use of these tools from an ethical perspective. Notably, we map out the many types of risks in the use of conversational chatbots in medicine to the principles of medical ethics. In doing so, we propose a framework for better understanding the effects of these chatbots on both patients and the medical field more broadly, with the hope of informing safe and appropriate future developments.