Carl Westin

144 Consistency and agreement in conflict resolution Study 2 results showed some interesting differences between trainees and ex- perienced controllers. The more experienced group had notably less solution vari- ability for solving the same conflict repeatedly. As such, experienced controllers appeared more in agreement on a preferred solution, especially in regards to the control problem classification. In contrast to trainees, all experienced controllers were found consistent according to the control problem classification and all, ex- cept one controller, agreed on the same control preference. Taken together, this sug- gests that the control problem classification best describes conflict solution among experienced controllers. Further research is required in order to determine whether the control problem classification is better than other classifications in quantifying controller expertise. 6-8-4 Sensitivity of consistency and agreement measures Results showed that consistency and agreement was restricted to the dichotomous higher-level decisions captured by DS 1 (type of classification) and DS 2 (aircraft choice) in each classification hierarchy. More specifically, our results indicate that controllers can be considered consistent in terms of their decision to interact with one or both aircraft (solution parameter), having one aircraft go in front or behind another (control problem), and their preference for achieving a specific solution geometry. In contrast to the solution parameter hierarchy, the first two decision stages of the control problem and solution geometry classifications imply which resolution type (DS 3) and direction (DS 4) to implement in order to achieve the overarching control or geometry preference and aircraft choice (DSs 1 and 2). Moreover, our analysis revealed that controllers are inconsistent in regards to the more detailed decision stages of the classification types. In regards to DS 3 (resolution type), DS 4 (direction), and DS 5 (directional value), the classifications proved too detailed and thus captured more noise than signal. Consistency did also not improve when considering expertise degree separately (Study 2). This yields the question whether the classification scales are good enough to capture consistency and agreement. Our conclusion is that the classifications are sufficiently sensitive for higher aggregated decision levels that capture the overall solution characteristics. For these levels, a move towards individual sensitive automation is sensible. 6-8-5 Defining and measuring consistency The three strategy classifications identified in this paper overlap in the sense that decision stages in respective hierarchies can be combined in several ways. Accord- ingly, some participants were found consistent according to more than one classifi- cation. In addition, participants were found to both agree and disagree depending