The Importance of Diverse User Goals When Designing an Automated COVID Risk Counselor

Abstract

The COVID-19 Pandemic has required everyone to understand and respond to the ever-evolving threats during daily life. The overall goal of our work is to develop a system based on an interactive conversational agent that functions as a ‘risk counselor’ by providing personalized information to people who have concerns about the risk of COVID infection when participating in daily situations. Such a system could also be useful for people who have concerns about other kinds of infection (e.g., HIV), and more broadly, risk in other domains (e.g., radon exposure in the home). As an initial step, we conducted MTurk surveys to elicit people’s concerns about COVID risk in daily situations. A key takeaway from this study is that when designing an agent, there needs to be a better framing of risk-related recommendations about daily situations, which may be particularly effective for educating people about the risk and encouraging actions that address risk. This framing should also address diverse user goals, including emotional responses, uncertainty, planning, and opinions about these situations.