Transparent systems, opaque results: a study on automation compliance and task performance.

Area of Science:

• Human-Computer Interaction
• Cognitive Psychology
• Maritime Safety

Background:

• Automated decision aids are increasingly used in safety-critical domains.
• Understanding factors influencing user compliance with these aids is crucial for effective implementation.
• Transparency and perceived reliability are key aspects of human-automation interaction.

Purpose of the Study:

• To investigate how different forms of transparency in an automated decision aid affect user compliance.
• To examine the influence of task difficulty and aid reliability on compliance and trust.
• To explore the relationship between trust and compliance in automated systems.

Main Methods:

• Two experiments were conducted using a simulated nautical collision avoidance task.
• Experiment 1 manipulated transparency through pre-task instructions about the aid's attributes.
• Experiment 2 manipulated transparency via confidence estimates and varied problem difficulty.

Main Results:

• Transparency in pre-task instructions positively influenced compliance and task performance.
• Confidence estimates did not improve compliance; difficult problems led to lower compliance.
• Lower compliance on difficult problems was mediated by reduced aid reliability and trust.
• Trust and compliance showed low correlations across both experiments.

Conclusions:

• The type of transparency implementation significantly impacts its effectiveness.
• Problem difficulty can decrease compliance by undermining perceived aid reliability and trust.
• Findings offer a framework for understanding human-automation interaction dynamics in complex tasks.