Carl Westin

3-6 Conclusions 57 3-6 Conclusions Our results support the hypothesis that controllers are more accepting of advisories that match their own solutions. In fact, conformal advisories were accepted nearly 33% more frequently than were nonconformal. Notice, however, that if 75% of conformal advisories were accepted, this means that 25% of conformal advisories were actually rejected. This result was interesting of itself, and in the end stood out among all others: of 256 conformal solutions (i.e., replays of controllers’ very own previous solutions), 64 (or 25%) were rejected by controllers. How is it that controllers would disagree with themselves one quarter of the time? One speculation is that individual controllers are simply inconsistent over time in the strategies they employ, and the given solution they might generate at any point in time. As a group, controllers rely on various strategies for workload regu- lation, 80 conflict detection 26 and conflict resolution, 29 and there are data to support the view that controllers are either homogeneous 138 or heterogeneous 139, 140 in their conflict judgment and resolution strategies. Previous research into controller strate- gies seems to have focused on inter-controller reliability , in other words, agree- ment. In terms of intra-controller reliability , in other words consistency, data are both sparse and unclear. This is an area that requires further research. Another speculation is that controllers were not necessarily biased against au- tomation, but rather against advice per se, be it from a colleague or from a presumed automation tool. Fundamental differences have been shown in how operators re- spond to, and develop trust in, human vs machine-generated advice, 72, 141 and it is interesting to speculate that our results would have been different had participants been instructed that solutions were coming from a colleague, as opposed to automa- tion.