Carl Westin

2-2 Resolving automation acceptance issues 23 2-2-2 Heuristic forms of automation Several research efforts have explored alternative approaches to decision aiding au- tomation that better acknowledge the psychological and behavioral variables that might color human decision-making. 18 One possibility is to model algorithms after human control strategies governed by heuristics. This would make the automation act in a manner consistent with how the human might act and thereby reduce the compatibility gap in decision-making strategies. Heuristics are cognitive shortcuts commonly applied to reduce cognitive work- load and increase efficiency in decision-making. 81 Heuristics can color our ex- pectations, however, and there are occasions when they can introduce systematic decision-making biases. This can result in sub-optimal strategies. 22 To eliminate such unwanted cognitive biases, heuristics can be formalized and refined. Benefits of such heuristic approaches lie in their ability to sort out noise in the perceived environment and consequently being able to better predict future states. 82 For example, satisficing is a cognitive heuristic that searches alternative solu- tions until an acceptable “good enough” solution is found. Satisficing is typically applied when an optimal solution cannot easily be determined. 20, 21, 83 Goodrich and Boer 84 took inspiration from this heuristic and hypothesized that car drivers would better understand and prefer adaptive cruise control automation that mim- icked skilled human task behavior for longitudinal vehicle control. They devel- oped a human-automation interaction framework based on multiple mental models that synthesized satisficing decision theory 21 with Rasmussen’s skill-, rule-, and knowledge-based taxonomy. 85 In a series of experiments, three human car driving skills were emulated by automation: active braking, speed regulation, and main- taining a desired distance (headway) to a preceding vehicle. For automation tuned after driver behavior and thresholds for the three skills, findings indicated improve- ments in the system’s performance evaluation, detection of the system’s functional limitations, returning to manual control, trust in the system, and overall safety. Similarly, researchers have shown that decision-making heuristics are com- monly used by air traffic controllers in CD&R, 23–28 and have advocated heuris- tic approaches to decision aiding automaton design. 29, 31–33, 61, 86 Controllers are believed to rely on cognitive processes akin to naturalistic decision-making when identifying and solving problems. 35, 87 While being able to rapidly interpret and act upon complex situations, controllers tend to struggle with providing strategy-based explanations underlying their decisions. 87, 88 Further, controllers are believed to de- velop and maintain conflict resolution heuristics in a “mental library”. 89 When a conflict is detected, the controller scans through the library for a suitable solution. Several research initiatives have attempted to mimic this decision-making pro- cess and have explored human heuristics-based approaches for decision aiding au-