Carl Westin

6-8 Discussion 145 on the classification used to define their problem-solving style. For example, while the control problem classification indicated contrasting problem-solving styles be- tween two participants, their different control actions resulted in the same spatial relationship between the two aircraft, as indicate by the solution geometry classifi- cation. As a consequence, the relevance of the consistency and agreement analysis in this study can be questioned. These findings were contradictory and surprising. However, they support generalization and assumptions of homogeneity among con- trollers in conflict resolution. The definition of what exactly constitutes a consistent problem-solving behavior is critical, if we are to develop automation that acknowledges and is sensitive to con- trollers’ problem-solving styles. Conflict solving is not a simple matter of choosing alternative A or B, it is considered more rich and complex. Yet our results suggest that consistency predominantly is limited to such dichotomous choices, even though a few participants were found consistent down to the fourth decision stage in the so- lution parameter hierarchy classification. The distinction between choosing a right vector, or a right vector with a speed increase may seem subtle but can be important. Future research is needed to address how to qualitatively distinguish solutions and define and measure consistency. 6-8-6 When to intervene Because a solution is valid only for a limited period of time, it is reasonable to ex- pect different solutions depending on when the conflict is solved. Results showed that intervention time varied greatly between both repetitions and controllers. Al- though participants’ intervention times were confounded by the conflict warning alert, the absence of any effects on how conflicts were solved is worth highlight- ing. This suggests that intervention time may be less important for how a conflict is solved. Similarly, previous research has found that differences in “look-ahead” time (5-12 minutes versus 8-14 minutes) do not influence controllers’ solutions. 35 Although further research is needed, this finding is relevant for ATC decision aid design since it suggests that specific conflict resolution advisories can be used for large time windows. The timing of advisories may, however, be important for other reasons, such as personalized to controller’s preferences for immediate interaction or to “wait and see,” and to avoid annoying and intrusive interruptions (i.e., advisory etiquette 114 ). 6-8-7 Controller strategies Several problem-solving styles similar to resolution strategies identified in previous literature were found. In particular the two strategies “vector behind” and “vec-