Common theories of operators' use of assistance systems often involve the concept of trust. It is assumed that trust guides operators' behaviour. They develop a certain level of trust towards the system based on the characteristics of the automation. This amount of trust should be appropriate in order to use the automation properly. However, some studies did not find empirical evidence for this mediator role of trust. A model which is taking these contradictory findings into account and which is able to predict the usage of an assistance system is missing so far. Based on a narrative analysis of five experimental studies investigating the effect of decision aids on human behavior, we have developed a model predicting the use of decision aids. The model does not only focus on the technical components and the level of trust, but also takes the perceived risk of the anticipated output and the perceived interpretability of the situation, i.e. uncertainty into account. The model assumes that risk and interpretability have a strong influence on using an assistance system. Furthermore it is assumed that they moderate the mediator role of trust. Practical implications are made, regarding the importance of multiple sources of information to reduce operators' uncertainty. Empirical studies will show the influence of interpretability and risk on operators' behaviour and operators' trust towards the system. (Author/publisher)
Abstract