Carl Westin

2-4 Strategic conformance 33 2-4-3 Encompassing process and product Elements of strategic conformance include both process (underlying strategy) and product (outcome or solution), which can be considered independently. The so- lution arrived at by automation may be conformal with the human, although both have relied on different strategies. In contrast, similar strategies can result in dif- ferent outcomes. Although conformance of both process and product can be dis- cussed in theory, we cannot look at product and conclusively determine process. Yet, this is what we do all the time in the real world when interacting with col- leagues, friends and others. We make inferences about underlying motivations and strategies based on the repeated exposures to outputs that can be observed. This inference engenders attitudes, feelings, and responses such as trust and acceptance. Research suggests that these interaction characteristics also apply when we interact with automation. 72, 120 These type of responses have been shown to positively in- fluence acceptance and trust in robots, 115, 116 computers, 118, 124 automated decision aids in flight simulation, 114 and autonomous cars. 125 We envision product conformance to be most valuable in situations that lack objective “gold standard” criteria for determining an unequivocal best or optimal solution. In such situations, the human is the best judge of decision quality. Strategic conformal automation can still be beneficial in these situations, primarily in terms of speed at which solutions are addressed, and in alleviating operator workload. More importantly, it should foster acceptance, trust and willingness to use the automation. Process conformance, however, may best apply in situations for which objective “gold standard” criteria are available. Automation using the same criteria as the human will likely be able to more consistently and accurately solve problems. As such, process conformance can more drastically benefit performance. Still, it would be conformal in that the automation would assess, weigh and combine criteria based on an understanding of the individual human process. Strategic conformance based solely on process could be problematic, because underlying strategies may be difficult to determine objectively, and therefore in- ferred primarily based on observable behavior. As an alternative, an individ- ual’s consistent problem-solving pattern can be used to determine that individual’s problem-solving style. If automation would be attuned to and capable of learning about an individual’s preferences, it could be developed to generate personalized advisories conformal to those preferences. This notion is captured in the research area of recommender systems, which attempt, based on a personalized user model, to predict individual user preferences and provide recommendations thereafter. In- formation about users’ preferences is often gathered through approaches relying on rating structures. Recommendations are typically based on the user’s past prefer- ences, or that of other users with similar profiles. 126