Presentation
A Rational Operator's Behavior Working with Diagnostic Automation with Fully Informative Automation Validity Information
DescriptionTo assess human responses to diagnostic automation, it is essential to define the optimal response a human operator should display. Meyer (2004) mathematically proved that the extreme responding strategy is the theoretically optimal response. A key assumption in the extreme responding strategy is that the operator’s response is binary, i.e., s/he either (blindly) follows or (blindly) ignores the diagnostic automation. The assumption, however, does not hold in many scenarios. Often, human operators can cross-check and obtain extra information, usually at a cost, to validate the output of the diagnostic automation before making a decision. In this work, we relax the binary assumption and derive the optimal strategy a rational human operator should use when automation validity information (AVI) is available. We calculate the expected utility of three responses: cross-checking (i.e., obtaining AVI), blind compliance, and blind ignorance. When AVI is fully informative, we identify a lower and upper bound. Within these bounds, the operator should always cross-check; outside of them, the extreme responding strategy is optimal.
Event Type
Creating AI that Works for People
Lecture
TimeThursday, October 16th11:30am - 11:50am CDT
LocationGrand B
Cognitive Engineering & Decision Making
Creating AI that Works for People: Human-Centered Innovation