Inhibitory Cognitive Control Allows Automated Advice to Improve Accuracy While Minimizing Misuse

Abstract

Humans increasingly use automated decision aids. However, environmental uncertainty means that automated advice can be incorrect, creating the potential for humans to act on incorrect advice or to disregard correct advice. We present a quantitative model of the cognitive process by which humans use automation when deciding whether aircraft would violate requirements for minimum separation. The model closely fitted the performance of 24 participants, who each made 2,400 conflict-detection decisions (conflict vs. nonconflict), either manually (with no assistance) or with the assistance of 90% reliable automation. When the decision aid was correct, conflict-detection accuracy improved, but when the decision aid was incorrect, accuracy and response time were impaired. The model indicated that participants integrated advice into their decision process by inhibiting evidence accumulation toward the task response that was incongruent with that advice, thereby ensuring that decisions could not be made solely on automated advice without first sampling information from the task environment.

Identifying Inefficient Strategies in Automation-Aided Signal Detection

“…The current models are also restricted to analysis of response accuracy. Models of response time (Strickland et al, 2021) and confidence (De Martino et al, 2013;Pleskac & Busemeyer, 2010) might provide further insight on the automation-use strategies suggested by our data.…”

Automated decision aids: When are they advisors and when do they take control of human decision making?

“…Given that Strickland et al (2021) did not previously find effects of automation on thresholds, we expected to replicate that. However, participants can control their thresholds to affect the relative speed versus accuracy of their responding (e.g., Boag, Strickland, Loft, et al, 2019), and tend to optimize how quickly they can complete an experiment while maintaining an acceptable level of accuracy (Hawkins et al, 2012).…”

“…The ATC task, depicted in Figure 2, was developed by Fothergill et al (2009) and was used by Strickland et al (2021). Participants viewed a 180 nautical miles (nmi) by 112.5 nmi sector of airspace, with a 10 nmi by 20 nmi scale fixed on the left of the display.…”

“…The Strickland et al (2021) model was tested in a simulated ATC conflict detection task in which participants were provided a 90% reliable decision aid recommending a classification (conflict or nonconflict). The decision aid improved conflict detection accuracy on trials where it was correct and decreased accuracy on trials where it was incorrect.…”

“…Although the Strickland et al (2021) model has the potential to identify the cognitive mechanisms by which increased trust in automation allows a decision aid to autonomously trigger decisions (i.e., excitation), to date the model has only implicated inhibition. This suggests that the decision aid in Strickland et al played the role of advisor, without potential to autonomously trigger decisions.…”