Carl Westin

1-3 Decision-making mismatches 5 concluded that the system was rejected in conflict situations, as it did not reflect controllers’ current way of managing conflicts. 9 Researchers have explored alternative approaches to CD&R decision aids that acknowledge the psychological and behavioral variables that might influence the controller’s conflict resolution strategies and solutions. 18 It is, therefore, worth investigating how similarities and differences between controller and automation problem-solving activities may affect the acceptance of decision aids. As such, the key problem this thesis addresses is therefore: Problem definition How to overcome controller acceptance issues of automated decision aids for conflict detection and resolution? 1-3 Decision-making mismatches Automated CD&R decision aids have generally been designed with a limited con- sideration for controllers’ individual decision-making processes or solution prefer- ences. Conflict-solving algorithms typically approach the environment in a dichoto- mous fashion, providing single, fixed, mathematically optimal solutions according to causal deterministic laws. 19 From a technology-centered perspective, this is not an issue: a system-generated optimal solution should be accepted and the controller should manage only by exception. However, one potential human performance problem is that an optimized (e.g., single vector) solution can hide the automation’s “reasoning” and paradoxically present a solution that the controller cannot easily evaluate. As automation becomes more advanced and assumes more of the “think- ing,” the controller’s interpretation and understanding of what the system is doing and why may become more critical. In contrast, psychology researchers argue that humans tend to approach problem-solving more heuristically (i.e., intuitively and by rule of thumb) and quickly settle for solutions that satisfice rather than optimize. 20–22 Analogously, research has shown that controllers commonly rely on heuristics for CD&R 23–28 and settle for a “good enough” conflict solution that works. 24, 29 Therefore heuris- tic approaches, as opposed to optimized algorithmic ones, have been advocated for human-centered CD&R decision aid design. 29–33 One example system is the Con- troller Resolution Assistant (CORA) tool intended as a CD&R decision aid for en- route controllers. 29, 34, 35 The algorithm is based on a template of controller heuris- tics in conflict resolution. This is achieved by constructing a library of controller strategies and identifying a set of “best” solutions (around four) that matches a ma- jority of controllers (investigations suggest that 80% or more is reasonable). 35 When detecting a conflict, the CORA algorithm provides the controller with a list of alter-