Automation Error Bias, Trust, and Dependence Behaviors in a Simulated Drone Collision Avoidance Task

Area of Science:

• Human-Computer Interaction
• Cognitive Psychology
• Robotics

Background:

• Prior research highlights asymmetrical effects of automation errors (misses, false alarms) on user compliance and reliance.
• The underlying reasons for these asymmetrical effects—perceptual salience versus trust calibration—remain unclear.

Purpose of the Study:

• To investigate how error biases in an automated decision aid influence trust and dependency behaviors.
• To examine the impact of miss-prone versus false-alarm-prone automation on user trust and reliance in a simulated drone collision avoidance task.

Main Methods:

• Sixty-eight participants engaged in a simulated drone monitoring task with an automated collision avoidance aid.
• The aid's error bias was manipulated (miss-prone vs. false-alarm-prone) with matched perceptual salience.
• Measures included automation trust ratings and dependency behaviors (compliance, reliance).

Main Results:

• Participants exhibited decreased trust across multiple factors when interacting with unreliable automation, regardless of error bias type.
• Compliance rates were higher with miss-prone automation compared to false-alarm-prone automation.
• Reliance rates showed an inverse pattern, being higher with false-alarm-prone automation.

Conclusions:

• Automation error bias systematically influences user compliance and reliance behaviors.
• Saliency-matched miss and false alarm errors degrade user trust, potentially hindering performance-based trust development.
• System designers must consider error type impacts on dependency to calibrate trust appropriately and enhance transparency.