Key to studying and assessing trust in human agent teams (HATs) is the ability to measure it. The predominant approach to assessing team emergent states and trust, more specifically, has been through self-report survey methodologies. However, self-reported trust measures suffer from several limitations including confounding with cognitive biases and social desirability (e.g. Arnold & Feldman, 1981; Taylor, 1961), inaccuracies due to retrospective assessments of abstract concepts (Podsakoff & Organ, 1986), assessment of trust as a static state rather than a dynamic process of emergence (Kozlowski, 2015), and the impracticality of asking members to pause activities to complete a survey. There is a clear need for innovative approaches to better capture trust, for both research and applied purposes. Recently, researchers have recommended and begun transitioning toward unobtrusive measurement methodologies such as physiological measures, event-based behavioral assessments, and language/communication (Azevedo-Sa et al., 2020; Hill, White, & Wallace, 2014; Marathe, Brewer, Kellihan, & Chaefer, 2020; Waldman, Wang, Stikic, Berka, & Korszen, 2015). Psychological measures, including subjective measures, have been correlated with more objective, unobtrusive measures of trust in HATs (Khalid, Helander, & Lin, 2021). For instance, physiological measures such as voice tone and pitch, facial expressions, heart rate, heart rate variability, and electrodermal activity have been shown to indicate trust (Khalid, Helander, & Lin, 2021; Schaefer et al., 2021). Furthermore, behavioral measures such as posture, eye fixations, allocating tasks to the autonomous agent, and manually controlling an agent have also been shown to correlate with trust (Schaefer et al., 2021; Khalid, Helander, & Lin, 2021). This paper will present an integrative review of the literature in this field and propose a theoretically-grounded, Unobtrusive Measurement Framework of Trust Dynamics in HATs that will more accurately, effectively, and practically capture trust in HATs than traditional measurement approaches.
Keywords
AGENT-BASED SIMULATION,AI,HUMAN PERFORMANCE,MEASURES,TEAM TRAINING
Additional Keywords
Unobtrusive measurement, trust, human-agent teams, team dynamics