Carl Westin

178 Conclusions sories. Conformal advisories were found to benefit both acceptance and agreement, as well as reduce the response time to advisories, both among experienced con- trollers and trainees. Notwithstanding the limited sample in the Source bias study, results provided valuable insights into controllers’ perceptions of automated ver- sus human advisers. In support of previous findings, the human adviser (presented as a controller) was perceived to provide solutions slightly safer, more efficient, and more similar to the controller’s own solution. Whilst no interaction was found between strategic conformance and interface transparency in the Automation trans- parency study, the findings do suggest that the problem-solving style definition (i.e., how an individual solves a particular problem), hence the conformal advisory, also depends on the human-machine interface used. As such, a conformal advisory when using interface A, may not be conformal with interface B. In addition to benefits observed in this thesis, there are numerous potential ben- efits of conformal automation that deserves attention in future research. Further studies could assess the use of strategic conformal automation in training, as a tool for learning operators to solve problems better. First gaining operators’ acceptance and trust during early interaction, then gradually change to suggest better solutions (that may at first be nonformal to the operator). Future research might explore alternative approaches to strategic conformal automation. As an alternative to ex- plicit advisories, conformal support can be facilitated by designing interfaces that allow operators to solve problems in their own preferred way. Ecological interfaces, for example, visualize the “objective truth” by revealing and showing the full con- straints affecting a situation. Importantly, such interfaces acknowledge that humans are heterogeneous in problem-solving, although the objective truth of a situation is homogeneous. This freedom may, however, not be practicable in all contexts. Future ATC CD&R, for example, is striving to relinquish tactical control from the controller altogether, using more advanced automation. Given the expected traffic growth, it is unlikely that, in the future, the controller will have time to consider conflicts in detail. In such contexts, providing explicit advisories conformal to the operator’s problem-solving style is a candidate approach for relinquishing problem-solving to the automation and freeing up the operator’s time and resources. To move the debate forward, a better understanding needs to be developed in regards to the suitability of personalized automation for different domains. The advancement in automation design has made it appropriate to question and challenge the development of “one-size-fits-all” systems, in favor of more person- alized applications. A natural progression of this thesis is thus the development of real-time personalized, conformal automation. There are, however, several chal- lenges, including strenuous technical ones in regards to the measurement and iden-