Presentation
A Dynamic Trust and Distrust Influence Metric that Predicts Team Trustworthiness and Affective Trust in Human Teams and Human-AI Teams
DescriptionThe study introduces influence as an information-theoretic measure of trust and distrust spread based in dynamical systems theory to capture the spread of trust and distrust in human-AI teams over time. It utilizes average mutual information to determine the degree to which joint team member actions influence system-level states over time. Forty-five three-member teams completed five 40-minute reconnaissance missions. Participants assumed the photographer role and worked with two confederate experimenters in navigator and pilot roles who portrayed either human or AI teammates. Trust and distrust were spread communicatively by the navigator as a between-subject condition and behaviorally by the pilot as a within-subject condition. Three influence time series measures for each teammate joint pair were calculated each mission and used in a series of repeated measures multiple regressions to predict individual performance and self-reported trust after each mission (i.e., team trustworthiness; cognitive trust and affective trust in each teammate). Team trustworthiness was predicted in the control condition and affective trust in the pilot was predicted in the communicative trust spreading condition. The results suggest that the influence measure is sensitive to behavioral and communicative spreading in HATs.
Contributors
Event Type
Lecture
TimeWednesday, October 15th2:10pm - 2:30pm CDT
LocationGrand A
Aerospace Systems
Human AI Robot Teaming (AI)