The Effect of Workload and Task Priority on Multitasking Performance and Reliance on Level 1 Explainable AI (XAI) Use

Area of Science:

• Human-Computer Interaction
• Cognitive Psychology
• Artificial Intelligence

Background:

• Operators in critical environments face multitasking challenges impacting performance.
• Explainable Artificial Intelligence (XAI) can support decision-making, but its use in multitasking is not well understood.
• Level 1 XAI provides basic perceptual information to aid operators.

Purpose of the Study:

• To examine how workload and task priority affect multitasking performance.
• To investigate operator reliance on Level 1 XAI systems in high-stakes scenarios.
• To understand the interplay between workload, task priority, and XAI utilization.

Main Methods:

• A within-subjects experiment with 30 participants in a simulated UAV command and control task.
• Manipulation of workload (low, medium, high) and AI-assisted task priority (low, high).
• Measurement of performance metrics including accuracy, AI reliance, and alert detection.

Main Results:

• Increased workload degraded performance on the AI-assisted task and heightened reliance on the AI system, particularly with low task priority.
• Task priority significantly influenced the use of AI explanations.
• Operators showed increased reliance on AI alerts under high workload but decreased alert verification.

Conclusions:

• Workload influences operator reliance on AI, necessitating careful calibration of AI trust in critical systems.
• Task priority is a key factor in how operators engage with AI explanations.
• Findings inform the design of AI systems for high-stakes environments to ensure appropriate AI reliance.