Carl Westin

82 Source bias effects 4-6 Conclusion The main question addressed in this study was whether operators are biased against human and automated decision aids differently, and to what extent this affects advice acceptance in relation to the aid’s advisory conformance. A real-time ATC simula- tion was used to investigate source bias effects of air traffic controllers’ (i.e., expert operators’) trust in a decision aid, and acceptance of its conflict resolution advice. Controllers were led to believe that advisories were derived from either another con- troller (human) or from an automated system (automation). In addition, controllers were unaware that each advisory was either conformal (individually matched) or nonconformal (based on a colleague’s deviating conflict solution style). Taken together, this limited study suggests that operators have different biases toward human and automated decision aids. Questionnaire trust ratings indicated a preference for the human source. As such, internal perceptions and preferences appear to differ between sources, However, neither source bias nor advisory confor- mance was found to significantly affect the acceptance, agreement, or response time to advisories in the real-time simulation. As such, the effect of these internal source biases does not appear to strongly influence the interaction with different sources in terms of objective metrics. For future research, researchers should differentiate between trust and reliance measured before and after interaction. In addition, researchers should differenti- ate between whole-system evaluations of trust and reliance, and derivative trust and reliance measured during a task and interaction with an aid (e.g., as measured by ac- ceptance or agreement with its advisories). In conclusion, the measures of reliance and trust during use of a decision aid reflect the performance of the aid in the con- text in which it is used. In contrast, whole-system evaluations of trust and reliance also consider other characteristics of an aid, such as its perceived competence and knowledge.