Modeling trust and its dynamics from physiological signals and embedded measures for operational human-autonomy...

Area of Science:

• Human-computer interaction
• Cognitive science
• Biomedical engineering

Background:

• Human-autonomy teaming is critical in operational settings like space missions and public safety.
• Accurate trust modeling is essential for effective human-autonomy collaboration.
• Current trust models are often descriptive, lacking predictive capabilities.

Purpose of the Study:

• To develop an objective, non-disruptive model for inferring and predicting human trust in autonomous systems.
• To capture the dynamic nature of trust during human-autonomy interaction.
• To advance the field of trust modeling with predictive capabilities.

Main Methods:

• Collected physiological data (ECG, respiration, EDA, EEG, fNIRS, eye-tracking) and interaction data from 12 participants in a simulated remote monitoring task.
• Used ordinary least squares regression to fit a model predicting subjective trust levels.
• Extracted and down-selected features from 2304 observations.

Main Results:

• The developed model achieved high accuracy with a Q² of 0.64.
• The model successfully captured rapid fluctuations in trust during the task.
• Physiological signals were effectively integrated into a predictive trust model.

Conclusions:

• The study demonstrates the feasibility of using bio-signals for predictive trust modeling in human-autonomy teaming.
• This approach offers a non-disruptive method for assessing trust, advancing beyond descriptive models.
• Future research should validate model performance with new participants.