Contacts:

• Tye Rattenbury (rattenbt -at- cs.berkeley.edu)
• John Canny (jfc -at- cs.berkeley.edu)

Brief Project Summary:

We are primarily concerned with functionalizing a definition of context. Following many findings within and outside Human Computer Interaction, we think human activity provides the proper framework for defining context and for identifying relevant contextual information. Accordingly, we are exploring methods for automatically detecting activity patterns.

We have built an automatic task support system called CAAD, for Context-Aware Activity Display. CAAD reliably detects patterns that correspond to people's stable working contexts by applying a tailored data mining algorithm to detailed computer logs. Preliminary experiments demonstrate that users find CAAD useful and easy to use.

Publications:

• Rattenbury, T. and Canny, J. "CAAD: An Automatic Task Support System," in Proceedings of SIGCHI. ACM Press, 2007.
• Canny, J. "GAP: A Factor Model for Discrete Data," in Proceedings of SIGIR. ACM Press, 2004.