Linking precursors of interpersonal trust to human-automation trust: An expanded typology and exploratory experiment

Abstract

ABSTRACT This study provides an initial experimental investigation of the extent to which well-known precursors of interpersonal trust (ability, benevolence, integrity, or ABI) will manifest when assessing trust between a human and a non-human referent (e.g. an automated aid). An additional motivation was the meta-analytic finding that the ABI model only explains about half of the variation in interpersonal trust. Based on a review of interpersonal and automation trust literatures, two additional precursors to trust – transparency and humanness – were identified and studied as exogenous variables (with A, B, and I analysed as explanatory mediators of their relationships to trust). In our experimental task, users interacted with an automated aid in decision-making scenarios to identify suspected insurgents. Results indicated that perceived humanness of the aid significantly correlated with trust in that aid (r = .364). This relationship was explained in part by perceptions of both ability and benevolence/integrity (unit-weighted average) of the aid; the latter finding suggesting that human-like intentionality attributed to the aid was a factor in automation trust. Perceived transparency also significantly correlated with trust (r = .464) although much of this relationship was explained by ability rather than benevolence/integrity. Aid reliability was also varied across the experiment. Interestingly, the explanatory power of benevolence/integrity increased when the aid’s reliability was lower, again suggesting human-like intentionality matters in automation trust models. Research and design considerations from these findings are noted.