Quantifying Human Trust in Controlling Robot Swarms: EEG-based Analysis and Classification.

Area of Science:

• Human-Computer Interaction
• Robotics
• Neuroscience

Background:

• Trust is crucial for human relationships but challenging in human-robot interactions.
• Direct control of robot swarms by humans requires understanding trust dynamics.
• Current methods for assessing trust in autonomous systems are limited.

Purpose of the Study:

• To investigate neural correlates of trust during human control of robot swarms.
• To develop a machine learning model for quantifying trust levels from brain activity.
• To explore the potential for adaptive robot swarm behavior based on trust.

Main Methods:

• Collected Electroencephalography (EEG) data from human subjects controlling robot swarms via joystick.
• Introduced noise into the system to induce distrust.
• Applied machine learning techniques to classify EEG correlates of trust.

Main Results:

• Identified distinct neural correlates of trust in human-robot swarm interaction.
• Achieved classification accuracy greater than 88% in discerning trust levels from EEG data.
• Demonstrated the feasibility of quantifying trust in real-time.

Conclusions:

• Neural correlates of trust are discernible in humans interacting with robot swarms.
• Machine learning can accurately quantify trust, facilitating human-robot teaming.
• This research is vital for deploying robot swarms in critical applications like exploration and defense.