Carl Westin

154 Discussion and recommendations Automation Transparency. The Automation transparency study sought to an- swer the degree to which controllers’ acceptance and understanding of resolution advisories depended on the transparency afforded by the interface. Two interface transparency levels were used: the baseline low level heading band SSD, and the high level triangle SSD which was a richer display showing more details of the traffic constraints. Simulation data did not indicate any interaction effects between automation transparency and advisory conformance on acceptance of performance. Results, however, indicated a preference for the more transparent interface, with controllers finding it more transparent and understandable. In addition, controllers’ working methods changed when using the triangle display, to more speed and com- bined clearances. In support of previous research, 121 results show that constraint- based interfaces can be used to guide the creation of more understandable interfaces, explaining the automation’s solution rationale. Transparency and strategic conformance can be seen as complementary in re- gards to objectives, but different in regards of methodology. The objectives of both argues that the human-automation relationship benefits from the automation facili- tating an increased understanding of how it operates, thinks, makes decisions, and acts. In terms of methodology, increased transparency is often achieved through ex- planations or by providing more (meta) information, for example through graphic- based visual approaches such as EID (Chapter 5). In contrast, strategic conformal automation argues that if the automation thinks like the operator, its advice can be tacitly understood. As such, lessening the need for increased interface transparency. Controller experience. Trust and credibility research has established that self- confidence (e.g., knowledge and ability) is closely tied to how trustworthy and cred- ible a support system is perceived. 15, 50, 68, 108, 160–164 Experienced operators may be reluctant towards novel automation because they have 1) high confidence in their own ability, 2) previously managed without support, 3) developed their own func- tional way of working, and 4) are satisfied with the current work practices. As hypothesized in this thesis, however, experienced controllers may be more accept- ing of conformal automation because resolution advisories, in fact, are their own. Conversely, novice operators can be expected to have lower self-confidence and be more amenable to others’ authority and expertise. Hence novices may be more embracing towards novel automation as they have 1) little routine, 2) are still learn- ing, 3) lack well-established practices, and 4) have not (yet) developed their own problem-solving styles. Novices’ performance was therefore expected to be more variable and less consistent. Results show that conformal advisories benefited both experienced controllers (First empirical study) and trainees (Automation transparency study), although ef- fects were less pronounced among trainees. Moreover, experienced controllers were