How Humans Comply With a (Potentially) Faulty Robot: Effects of Multidimensional Transparency

Abstract

This article deals with how a human operator follows a request made by a robot in an industrial context. This robot is potentially myopic, i.e., its request could be based on partial and limited information. Therefore, it could possibly be faulty. This exploratory study aims to analyze whether multidimensional agent <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">transparency</i> may have an effect on two drivers of compliance identified in the literature: trust in signal and risk perception. In this experiment, we manipulated different agent transparency levels combined with two dimensions of agent transparency: robot-TO-human transparency (rTOh) and robot-OF-human transparency (rOFh). Results mainly show that adding rOFh to rTOh transparency changes human compliance with the agent and moderates both trust and risk perception. Moreover, task performance and completion time are shown to vary according to the different transparency levels. Our results show that transparency has no effect on mental workload. From a methodological perspective, this article shows the importance of distinguishing the different types of information about which a robot can be transparent, especially the combination of rTOh and rOFh transparencies. From a practical point of view, the article shows that the agent transparency framework needs to be considered carefully when designing human-robot teaming in the context of Industry 4.0 in the case of robots that may be unreliable due to their myopia.