On using existing time-use study data for ubiquitous computing applications

Abstract

Governments and commercial institutions have conducted detailed time-use studies for several decades. In these studies, participants give a detailed record of their activities, locations, and other data over a day, week, or longer period. We show in this paper how data from these studies may be used to improve ubiquitous and mobile applications. In particular, the large numbers of participants in these studies (often in the tens of thousands), gives a rich source of information to serve as prior probabilities for activity-inference. Furthermore, the representations of activities developed for these studies are the product of much investigation, and could serve as references for ubicomp applications. Finally, the data within the studies can show which variables are particularly effective at predicting activity.