Evaluating the Effects of Displaying Uncertainty in Context-Aware Applications

Many context aware systems assume that the context information they use is highly accurate. In reality, however, perfect and reliable context information is hard if not impossible to obtain. Several researchers have therefore argued that proper feedback such as monitor and control mechanisms have to be employed in order to make context aware systems applicable and useable in scenarios of realistic complexity. As of today, those feedback mechanisms are difficult to compare since they are too rarely evaluated. In this paper we propose and evaluate a simple but effective feedback mechanism for context aware systems. The idea is to explicitly display the uncertainty inherent in the context information and to leverage from the human ability to deal well with uncertain information. In order to evaluate the effectiveness of this feedback mechanism the paper describes two user studies which mimic a ubiquitous memory aid. By changing the quality, respectively the uncertainty of context recognition, the experiments show that human performance in a memory task is increased by explicitly displaying uncertainty information. Finally, we discuss implications of these experiments for today’s context-aware systems.