Speaker: Matt Kay
, University of Washington
Title: Designing for user-facing uncertainty in everyday sensing and prediction
We are increasingly exposed to sensing and prediction in our daily lives (“how many steps did take today?”, “how long until my bus shows up?”, “how much do I weigh?”). Uncertainty is both inherent to these systems and usually poorly communicated—or not communicated at all. However, understanding uncertainty is necessary to make informed decisions from data: if my scale says I am a pound heavier today than yesterday, how does that compare to the expected variance of my weight? If my bus is predicted to arrive 10 minutes from now, what is the chance the bus instead shows up early, in 5 minutes—and therefore, do I have time to get a coffee? In this talk, I overview several projects I have undertaken to investigate how people deal with uncertainty in their everyday data and how we can build systems that convey uncertainty in a way they can understand. I will touch upon several domains, including weight tracking, sleep quality, and realtime transit prediction. I argue that how people interpret their data and what goals they have should inform the way that we communicate results from our models, which in turn determines what models we must develop and use in the first place. As we push more sensing and prediction into people’s everyday lives, we must consider carefully how to communicate estimates that people can actually use to make informed decisions.
Matthew Kay is a PhD Candidate in Computer Science & Engineering at the University of Washington. His research in human-computer interaction (HCI) spans many application domains, from personal health informatics, to communicating uncertainty in everyday sensing and prediction, to advancing the state of statistical methods in HCI. He draws upon qualitative and quantitative methods across HCI, information visualization, and statistics. He holds a Master's and a Bachelor's in Computer Science (minor in Fine Art) from the University of Waterloo.