This week we’ll be at MobileHCI’09 in Bonn, presenting a poster on our research within the Diadem project. One of the main goals of the Diadem project is to detect potentially hazardous airborne pollutants in urban-industrial areas using input from both a distributed sensor network and people through their mobile phones. In the proposed interaction model, a semi-autonomous system will use sensor data to detect abnormal situations, while people in the affected area will be requested by a mobile service to report additional observations, such as chemical smells (which may not be the easiest to describe).
This raises quite some interesting issues. Not only should such a system be capable of communicating via a wide variety of mobile devices; it especially needs to communicate with different people and the wide variety of situations they might be in. The system’s goals in gathering information might differ from users’ own, immediate goals. On the one hand, the system requires unbiased information from users to determine the likelihood and location of an incident. On the other hand, users would like to get information as well, express concerns, complain about unfavourable smells, or receive instructions in the (unlikely) event of a hazardous incident. In many cases, detected anomalies will not be indicative of a serious problem, but the system will have interrupted users anyway. Such a system needs to build a long-term relationship that motivates users to provide unbiased information. It should also react in a socially acceptable manner that takes into account the nuisance of interruptions, users’ other activities and their emotional state.
The project is still in its early stages, but there’s a lot going on already. This summer we’ve completed our first lab study investigating the effects of social system behaviour on user trust and behaviour. Our first findings indicate that getting social behaviour of a system wrong can have serious consequences on people’s reactions to system advice. The results of another study, investigating communicating with the general public during crisis situations, indicate that adaptation to the user’s affective state and information processing style can significantly improve interaction. In the mean time, we’re also conducting a longitudinal field study collecting smell descriptions and testing a first prototype of an intelligent mobile (phone) agent and its ‘stench classification dialogue’ (for which we might need a somewhat more charming working title ;)). Beyond the interaction scenario sketched above, we’re also developing other interaction concepts that utilise ‘semi-free’ sources of information (information offered on the users’ own initiative, social network status updates).
The MobileHCI paper can be found here. Do come round our poster if you happen to be at MobileHCI yourself!