The Role of Behavioral Anthropomorphism in Human-Automation Trust Calibration

Measuring and Understanding Trust Calibrations for Automated Systems: A Survey of the State-Of-The-Art and Future Directions

“…The remaining tasks (such as income prediction, number pattern discovery, pothole detection, prescription screening, classification of plants and animals, playing a game, prize prediction) involve fewer risks and might be less suitable for triggering trust. However, some studies increased vulnerabilities by introducing gamification strategies in which participants were (financially) incentivized to perform well [10,55] Interestingly, we noticed that a large domain, medicine/healthcare, commonly characterized as a high-stakes domain, was represented by only two papers. As it is important to learn more about trust in high-stakes domains, this is unfortunate.…”

“…Most studies manipulated the system's capabilities. The most common strategy to achieve this was to vary the levels of system reliability (see, e.g., [4,10,14,15,20,26,27,55,69,81,82,89,94,104,107,117,133,137,139,140,143]).…”

“…In addition to compliance and task performance, other performance measures were, for example, error rates and response times [81]. Most often, we found, however, that researchers used a combination of both self-report measures and behavioral measures (see, e.g., [4,14,19,26,47,50,55,61,84,94,117,130,139,140]). In two studies, researchers also added qualitative questionnaires to advance their quantitative trust assessments [84,90].…”

“…In many cases, trust measures were accompanied by subconstructs of trust such as perceptions of system accuracy [90,143], predictability [68,73], and reliability [15,68,107,133], or trust related measures such as perceived usability [4], frustration [4], perceived humanness [26,27,55,58], perceived intelligence [27], workload [58], and understanding of the system [132].…”

“…A third way to measure the trust calibration was to measure over-, under-, and appropriate trust as the behavioral deviation (reliance) from the ideal behavior [27,50,55,104,110,132,137]. For example, Jensen et al [55] asked participants to identify dangerous vehicles in a set of images. Participants could manually check the images or let an automated aid do it for them.…”