| The importance of understanding and supporting operators' judgment has been recognized in a variety of contexts. Automated decision aids have been developed to assist people in this respect. In doing so, operators' trust in automated systems has been discussed as a critical component affecting operators' performance. This research investigated how people make judgments in a complex environment when they are supported with automated decision aids, particularly with respect to how different characteristics of automated decision aids affect judgments, including the aspects that may contribute to decrease the level of trust in such systems.; A theoretical model within the Lens Model was developed to model a situation of multiple judgment agents, called Hybrid Lens Model. This model measures the effect of one agent's judgment on the other's using a correlation model. Accordingly, the Lens Model parameters were redefined to represent this effect in addition to the cooperative judgment performance. Within the context used in this research, the model measured the effect of an automated decision aid on the operator's judgment performance, where automated decision aids varied in two aspects: validity and reliability. A third factor, call understandability, was also tested. Lens Model based feedback regarding the automated decision aid's performance was provided.; For the study, participants performed an aircraft identification task for two consecutive days. While performing the task without the aid on the first day, participants performed the task with the aid, which varied in terms of its reliability and validity, and with the aid providing the feedback information in addition to the aid's estimates. Participants' performance was analyzed in terms of overall performance, Lens Model parameters, and subjective ratings of trust.; Results showed significant impacts of characteristics of the automated decision aid on judgment performance. Participants performing the identification task with support from the automated decision aid showed a range of performance depending on the quality of the automated decision aid's estimate. Participants' performance decreased as the level of the automated decision aid's quality decreased. The results showed that participants with the additional information performed the task better than those without it in every aspect of the Lens Model and related parameters. |
