Using fNIRS to Identify Transparency- and Reliability-Sensitive Markers of Trust Across Multiple Timescales in...

Area of Science:

• Human-computer interaction
• Neuroergonomics
• Social Psychology

Background:

• Intelligent agents are becoming teammates, requiring calibrated human expectations for optimal performance.
• Trust is a key factor in human-agent interactions, influencing reliance and utilization.
• Neurophysiological monitoring, particularly functional near-infrared spectroscopy (fNIRS), shows potential for measuring trust.

Purpose of the Study:

• To investigate the use of fNIRS for continuously monitoring trust in artificial agents within human-agent teams.
• To examine the impact of agent reliability and transparency on trust and team dynamics.
• To identify brain regions associated with trust assessment and mental demand in human-agent collaboration.

Main Methods:

• Developed a collaborative task for teams of 2 humans and 1 agent.
• Manipulated agent reliability and transparency.
• Collected neurophysiological data using fNIRS, alongside measures of trust, mental demand, team processes, and affect.

Main Results:

• Agent transparency and reliability significantly influenced trust.
• Reduced agent communication negatively impacted interpersonal trust and team cohesion.
• fNIRS data successfully predicted individual trust decisions up to 15 seconds in advance.
• Identified specific brain activations (dorsal medial prefrontal cortex, dorsal lateral prefrontal cortex, frontopolar cortex) linked to agent transparency and mental demand.

Conclusions:

• fNIRS is a viable tool for real-time trust monitoring in human-agent teams.
• Understanding brain activity can help calibrate human expectations and optimize agent collaboration.
• Future research can develop intelligent systems for real-time trust modulation to enhance team performance.