Activity modelling for low-intention interaction

When modelling user interactions, we normally assume that the user is acting with intention: some very explicit such as opening a valve in a nuclear power station, others more tacit, hardly needing any thought, for example tipping a tablet to turn a page. However, there are also a range of system behaviours that make use of unintentional user actions, where the user acts, but the system decides that the action has meaning, and how to make use of that meaning. Again, these may operate on a variety of levels from ‘incidental interactions’, which operate entirely without the user realising, perhaps subtle changes in search results based on past activity, to more ‘expected interactions’ such as automatic doors that open as you approach. For intentional interaction, there is long-standing advice—making sure that the user can work out what controls do, where information is, interpret the available information, receive feedback on actions—and also long-standing modelling techniques. Low-intention interactions, where the system has more autonomy, require different design strategies and modelling techniques. This chapter presents early steps in this direction. Crucial to this is the notion of two tasks: the sensed task, which the system monitors to gain information and the supported task, which the system augments or aids. First, this chapter demonstrates and develops techniques in the retrospective modelling of a familiar low-intention interaction system, car courtesy lights. These techniques are then applied proactively in the design of a community public display, which is now deployed and in everyday use.