The flexibility to search out clear, related, and personalised well being info is a cornerstone of empowerment for medical sufferers. But, navigating the world of on-line well being info is commonly a complicated, overwhelming, and impersonal expertise. We’re met with a flood of generic info that doesn’t account for our distinctive context, and it may be troublesome to know what particulars are related.
Massive language fashions (LLMs) have the potential to make this info extra accessible and tailor-made. Nevertheless, many AI instruments as we speak act as passive “question-answerers” — they supply a single, complete reply to an preliminary question. However this is not how an knowledgeable, like a physician, helps somebody navigate a posh matter. A well being skilled would not simply present a lecture; they ask clarifying questions to know the complete image, uncover an individual’s objectives, and information them by way of the data maze. Although this context-seeking is vital, it is a important design problem for AI.
In “In the direction of Higher Well being Conversations: The Advantages of Context-Searching for”, we describe how we designed and examined our “Wayfinding AI”, an early-stage analysis prototype, based mostly on Gemini, that explores a brand new strategy. Our elementary thesis is that by proactively asking clarifying questions, an AI agent can higher uncover a person’s wants, information them in articulating their issues, and supply extra useful, tailor-made info. In a collection of 4 mixed-method person expertise research with a complete of 163 contributors, we examined how individuals work together with AI for his or her well being questions, and we iteratively designed an agent that customers discovered to be considerably extra useful, related, and tailor-made to their wants than a baseline AI agent.