Carl Westin

170 Discussion and recommendations erator’s preferred problem-solving style, 2) be able to acquire an understanding of the operator’s reasoning in relation to the situation at hand, and 3) determine a so- lution for the given situation that matches this style. This capability implies that the system also can evaluate the suitability of a solution, and therefore be able to ex- plain why the conformal solution is considered inappropriate, and argue for a more appropriate, nonconformal, course of action. 7-4-7 Domains benefiting from conformal automation Although this thesis has been restricted to ATC CD&R, conformal automation has potential applications in any domain that contains automated systems. Directly re- lated applications can be found in other transportation domains that deal with sim- ilar contexts of collision avoidance, such as automation used in vehicle cockpits or traffic monitoring services. More broadly, however, strategic conformance does not relate so much to the task or problem at hand, but the behavior of the automation and the degree to which it complies with the human preferred way of working. In this perspective, conformal automation implies a more holistic design philosophy that strives to harmonize the automated system’s behavior to that of the human operator. 7-4-8 Homogeneity versus heterogeneity in automation design This thesis shows that controllers are more diverse than alike in conflict solving. In light of this, a relevant question is whether to strive for heterogeneity or homogene- ity in automation design. That is, should we design for the individual user, or for the population? The question echoes that of Hopkin [235, p. 81]: “Should system adaptability be viewed primarily as a means of encouraging individual differences between controllers or as a means of preventing them, since it has the potential for both?” Traditionally, design has favored the homogeneous view. Compared to automation, the human has been identified as exceptional in cre- ative thinking, identifying new solutions to problems, and the ability to adapt to its changing surrounding (e.g., Fitts list 283 ). Yet, systems’ interaction models and problem-solving algorithms tend to be fixed and narrow, instead appealing to the human strength of adapting to the system. In addition, these models and algorithms typically require that operators are trained homogeneously, approaching problems and solving them in line with the system’s rationale. In a way, this approach neu- tralizes the human variability, flexibility, and creativity. The argument in favor of heterogeneity (i.e., personalization), is that automation sensitive to the user’s prefer- ences and abilities can benefit acceptance, performance, enjoyment, and teamwork with the automated agent.