Carl Westin

156 Discussion and recommendations post hoc analysis of prequel data (i.e., controllers’ manual conflict solutions) from real-time simulations gathered in the three other studies (illustrated by the arrows feeding from the prequel simulations into the Consistency study). Data from the three real-time simulations were analyzed separately in two studies depending on whether the conflict was biased, or not (Study 1 and Study 2, respectively). To begin with, a framework consisting of three strategy classifications was cre- ated that describe different conflict-solving objectives. The solution parameter hier- archy argues that controllers determine solutions based on whether to interact with one or both aircraft. The control problem classification argues that controllers view the conflict as a control problem, solved by either clearing the controlled aircraft ahead or behind the other (by vector, speed, or combination thereof). Finally, the so- lution geometry classification argues that the resulting spatial relationship between the conflicting aircraft is the main driver for how the conflict is solved. The post hoc analysis confirmed a limited agreement between controllers, while all controllers were found consistent. Consistency was, however, primarily re- stricted to the highest two decision stages of each classification, such as vectoring the controlled aircraft behind or in front, or interacting with one or both aircraft. Below these high-level decision stages, consistency decreased with only a few con- trollers found consistent in regards to resolution type (decision stage 3: heading, speed, or combined) and direction (decision stage 4: right or left, increase or de- crease, or combinations thereof). Taken together, this suggests that controllers generally can be considered consis- tent in terms of high-level goals, but inconsistent in terms of specific maneuvers im- plemented. This provides an explanation for why controllers considered themselves inconsistent in questionnaire responses. Similarly, Inoue et al. observed diversity in conflict solutions among controllers, although they adhered to the same strategy. 247 In a study attempting to identify controller strategies by means of machine-learning techniques, Regtuit found that, with increasing variability in conflict solving, con- sistency could only be established for the two main strategies of either vectoring ahead or behind the other aircraft. 261 On the other hand, some inconsistency can be expected given the flexibility in the world, the flexibility and creativity in hu- man problem-solving, and the influence of human emotional states (e.g., tiredness, mood). As such, controllers can be expected to solve an identical problem differ- ently over time, and they may not like the solution once they see it replayed. Contrary to expectations, consistency and agreement were overall not higher for biased conflicts in Study 1 (i.e., data from First empirical study). However, while the solution geometry classification was favored (higher consistency and agreement) in Study 1, the control problem classification was favored in Study 2. This sug- gests that controllers’ solution variability did not lessen with biased conflicts. But