Near-Perfect Automation: Investigating Performance, Trust, and Visual Attention Allocation

Area of Science:

• Human-Computer Interaction
• Cognitive Psychology
• Automation Systems

Background:

• Near-perfect automated systems are increasingly deployed in high-stakes environments.
• Research on human performance, trust, and attention with such systems is limited.

Purpose of the Study:

• To assess human performance, trust, and visual attention during the monitoring of a near-perfect automated system.
• To quantify the subjective experience and attentional cost for operators.

Main Methods:

• Seventy-three participants performed a supervisory control task monitoring three search feeds.
• The automated system had 100% reliability except for one miss and one false alarm.
• Eye-tracking and subjective trust data were collected.

Main Results:

• Only 34% detected the automation miss, and 67% detected the false alarm.
• Subjective trust decreased upon detecting failures and increased when failures were missed.
• Failure detection correlated with altered scan patterns and increased attention to specific feeds.

Conclusions:

• Human limitations in monitoring near-perfect automation necessitate reevaluating operator roles.
• Individual operator states and attentional costs must be understood and designed for.
• Real-time, individualized operator assistance is needed for effective human-automation interaction.