Carl Westin

Summary iii several critical control parameters of the CD&R problem. It was used in all simu- lations as part of this thesis, and appeared when an aircraft was selected. As such, simulations represented a futuristic environment, different from current ATC opera- tions, with more traffic present, datalink communication between the controller and pilots, and the SSD supporting controllers in CD&R. Furthermore, simulations and conflict solving were restricted to the horizontal plane. This thesis reports on three human-in-the-loop studies. In addition, simula- tion data from these were analyzed post hoc in a fourth study. The first empirical study culminated in a large-scale real-time simulation with sixteen experienced con- trollers. The study varied strategic conformance (i.e., conformal or nonconformal resolution advisories) in addition to the system’s authority (LOA: management by consent or management by exception) and task complexity (high or low). Con- trollers accepted conformal advisories (i.e., advisories based on their own unique conflict solving style) more often, gave them higher agreement ratings, and re- sponded faster, than with nonconformal advisories based on a colleague’s contrast- ing but still workable and safe conflict solution. In 25% of cases, however, con- trollers disagreed with their own conformal advisories. The other two human-in-the-loop studies, and the post hoc analysis study, were conducted to further explore plausible causes for the observed disagreement, to- gether with other outstanding questions, derived in the first study. The two human- in-the-loop studies replicated the experimental approach used in the first study, with minor refinements. The Source bias study investigated differences in controllers’ acceptance of, and trust in, a conflict resolution aid, based on the presumed source of that advice. Five experienced controllers participated in a real-time simulation that varied strategic conformance together with the advisory source, presented to originate from either a human or automated source. While questionnaire responses indicated a slight preference for the human adviser, simulation results did not. The Automation transparency study investigated effects of interface trans- parency and strategic conformance on controllers’ acceptance and understanding of advisories. Nine controller trainees participated in a real-time simulation. Two levels were used, with the heading band SSD representing low transparency, and triangle SSD representing high transparency in that it provided more meaningful information on the relationship between conflicting aircraft. Results showed that the more transparent triangle SSD was better understood. Although no interaction effects between conformance and transparency were found, conformal advisories were accepted slightly more often than nonconformal advisories, supporting results from the first study. Moreover, when using the triangle SSD, conflicts were more often solved by speed and combinations of heading and speed. This indicates that