Abstract:
We propose that automation of user tasks can only be properly addressed as a particular problem of adaptation in human-computer interaction. We further claim that to achieve such automation, the ability to define, edit, and infer from models of user interfaces is needed. We introduce The Mecano Project, a modelbased interface development environment that allows precisely such capabilities. Finally, we describe how some traditional approaches to user-task automation, including machine learning techniques and programming by demonstration, can be incorporated into Mecano and have a higher potential of being successful.