Carl Westin

110 Automation transparency effects advisory: for example, this solution is more efficient because it takes aircraft A closer to the destination. A ‘tradeoff’ explanation would contrast alternatives and guides selection: for example, an alternative would be to vector the other aircraft, but it requires a larger deviation. Furthermore, although the objective of EID is to visualize the affordances for activity in relation to a particular work domain, it could perhaps be applied more specifically for visualizing a decision support system’s reasoning underlying recom- mendations and solution advisories. Across many domains, such specific solution advice has traditionally been provided in text-based formats. For instance, studies exploring ATC decision aids for conflict resolution usually provide the controller with a list of solutions in a tabular text format. 31, 35, 234 Like Woods et al. 199 note, the process underlying the generation and prioritization of text-based solutions is opaque to the controller, making it difficult for the controller to comprehend the solution(s). The interpretation uncertainties can lead to discrepancies between how the controller perceives the system’s suggested solution, and the solution preferred by the controller. 155 Furthermore, a list can encourage the operator to consider all alternatives before making a choice. Using text-based explanations to support an advisory (e.g., as advocated in recommender systems research) is not practicable in many domains given time pressures and workload demands. As an alternative to text, a rationale underlying the system’s advice can be provided graphically in the interface, by visually integrating information directly with advisories. Further research should investigate whether EID can be used for purposes of providing such explanations, for example by visualizing which constraints and complex relation- ships that have been considered when determining the suggested solution. 5-7-4 Transparency and conformance Results showed that participants were more likely to use speed and combined solu- tions to solve the conflict when using the TRI SSD. Although this was a desirable and not surprising effect attributed to the increased interface transparency, it implies that conformance also changes with the transparency. The conformance manipu- lation, however, was based on solutions made when using the HB SSD, for which speed and combined solutions were less common. As such, a heading solution may have been conformal when using the HB SSD, but not when using the TRI SSD. In part, this can explain why participants’ agreement was higher for nonconformal advisories than conformal advisories in the TRI SSD condition. The TRI SSD may have improved participants’ understanding of nonconformal advisories and guided their attention to solutions they did not think of. Moreover, the conformal advi- sories, based on solutions when using the HB SSD in the prequel simulation, may have appeared impractical when using the TRI SSD.