On the Dual Nature of Transparency and Reliability: Rethinking Factors that Shape Trust in Automation

摘要

Prior literature has found that increasing system reliability and transparency can positively impact operators’ trust of automated systems; however, these factors are typically confounded. In the present study, we separated these factors by manipulating different stages of automation. Participants engaged in a simulated coffee manufacturing task using an interface with differing levels of reliability (65% or 95%) and transparency (one line or multiple lines of system display). The Human Computer Trust Scale (HCTS) and the Trust in Automated Systems Scale (TAS) were used to measure trust. When examining scores on TAS items with a positive-valence, we novelly observed that transparency interacted with reliability, such that high transparency and low reliability negatively impacted trust in the system. Alternatively, trust was not negatively affected by poor reliability when transparency was low, due to trivial cost of corrective behaviors that compensated for poor reliability and lack of system history understanding by the operators.