The case for activity models
In this position paper, we argue that user modeling systems would benefit by explicitly modeling user activities. We give examples where systems without activity inference fail to perform as desired, and briefly describe both benefits (combining data from multiple sources, learning from multiple users, improving code and technique reuse, and collecting user feedback) and costs (system complexity, development time, and inference limitations).
Partridge, K.; Brdiczka, O. The case for activity models. Pervasive User Modeling and Personalization 2010 (PUMP); 2010 June 20; Big Island, Hawaii.