Towards Improving Trust in Context-Aware Systems Displaying System Confidence

Antifakos, S., Kern, N., Schiele, B., & Schwaninger, A. (2005). Towards improving trust in context aware systems by displaying system confidence. ACM International Conference Proceeding Series, 111, 9-14.

The principal contribution of this paper is a study in which we show the effect of displaying system confidence on user trust. The proposal is based on the fact that users are used to and highly successful in dealing with unreliable and uncertain information throughout their daily lives. The authors propose to display system reliability explicitly and leverage from the user’s ability to choose the appropriate action.

Our experiments show that when system confidence is displayed, users more often rely on the system. This suggests an increase of the user’s trust in the system.

Providing a feedback mechanism seems mandatory in the context of ubiquitous computing. Modeling uncertainties and advanced
inference mechanisms might not be enough. According to Bellotti and Edwards, context systems cannot be designed simply to act on our behalf. Rather they propose that those systems will have to defer to users in an efficient and non-obtrusive way.

Similarly, Bubb-Lewis and Scerbo [Getting to know you: Human-computer communication in adaptive automation] argue that the only way of reducing uncertainty is by exchanging information between the automatic system and the human user.

Context-aware and automatic system seem to provide similar feature. There are already some context-aware and automatic systems that propose the use of different feedback mechanisms. However, their effects on user’s trust and system usability have not been evaluated.

Reference to read:
B. Muir. Trust in automation: Part i. theoretical issues in the study of trust and human intervention in automated systems. Ergonomics, 37(11):1905–1922, 1994.

Relation to my thesis: It would be interesting to have a group perform a collaborative task and evaluate if displaying the ubiquitous environment reliability improves that performance. Atifakos et al. fail to mention the tradeoff between the cognitive load for displaying uncertainty and the added value that it provides. Evaluating the Effects of Displaying Uncertainty in Context-Aware Applications:

One issue to be considered in future work is the tradeoff between the cognitive load, which displaying uncertainty information causes, and the added value that it provides.

I follow Bellotti and Edwards (for context-aware) and Bubb-Lewis and Scerbo (for automatic system) who conclude that systems cannot be designed simply to act on our behalf. A challenge is to define what “efficient and non-obtrusive way” to communicate uncertainty, and nurture trust in the system. Trust and handling uncertainty are tightly related. There are many methods of presenting feedback in an effort towards making systems more predictable. Atifakos et al. show one way to evaluate them.