The Ambivalent Effect of Transparency on Trust in Robo-advisors: An Experimental Investigation

Rühr Alexander, Berger Benedikt, Hess Thomas

Abstract

Technological advancements have led to increasingly complex and autonomous algorithmic decision support and automation, decreasing system comprehensibility and user trust. These developments are well represented by robo-advisors in investment management. A potential lever to counter the loss of trust is a more transparent decisionmaking basis of such autonomous systems. Yet, the effect of transparency on trust may be ambivalent as the quality revealed by transparency may vary, potentially enhancing or reducing trust. To shed light on the (joint) effect of transparency and quality on trust in robo-advisors, we conduct an experiment in which participants interact with a manipulated robo-advisor representation. We find strong positive effects of transparency and quality on trust and a negative interaction revealing their substitutive capacity. Surprisingly, transparency proves particularly trust-enhancing in cases of low quality. Based on these findings, we derive theoretical and practical implications enhancing our understanding of interactions with increasingly autonomous decision support.