Master’s Thesis Defense Meeting

Date: July 14th, 2022
Time: 11:00 AM

Location: https://gatech.zoom.us/j/96776537555

Advisor:
Jamie Gorman, Ph.D. (Georgia Tech)
 
Thesis Committee Members: 

Jamie Gorman, Ph.D. (Georgia Tech)
Bruce Walker, Ph.D. (Georgia Tech) 

Christopher Wiese, Ph.D. (Georgia Tech) 

Title: Developing Objective Communication-based Measures of Trust for Human-Autonomy Teams 

Abstract: 

As artificial intelligence capabilities have improved, humans have begun teaming with autonomous agents that have the capability to communicate and make intelligent decisions that are adaptable to task situations. Trust is a core component of human-human and human-autonomy team (HAT) interaction. As with all-human teams, the amount of trust held within a HAT will impact the team’s ability to perform effectively and achieve its goals. A recent theoretical framework, distributed dynamic team trust (D2T2; Huang et al., 2021), relates trust, team interaction measures, and team performance in HATs and has called for interaction-based measures of trust that go beyond traditional questionnaire-based approaches to measure the dynamics of trust in real-time. In this research, these relationships are examined by investigating HAT interaction communication-based measures (ICM; amount, frequency, affect, and “pushing” vs. “pulling” of information between team members) as a mechanism for D2T2 and tested for predictive validity using questionnaire-based trust measures as well as team performance in a three-team member remotely-piloted aerial system (RPAS) HAT synthetic task. Results suggest that ICM can be used as a measure for team performance in real-time. Specifically, ICM was a significant predictor of team performance and not trust, and trust was not a significant predictor of team performance. Exploratory factor analyses of the trust questionnaire items discovered clear differences in how human teammates characterize trust in all-human teams and HATs. Specifically for HATs, interpersonal and technical factors associated with trust in autonomous agents were found as a result of dynamic exposure to the autonomous agent by distinct stakeholders through communication. These findings confirmed the underlying theory behind D2T2 as a framework that includes both interpersonal and technical factors related to trust in HAT along a dynamic timeline among different types of stakeholders. The findings provide some insight into the dynamic nature of trust, but continued research to discover interactive measures of trust is necessary.