Can Explainable AI Foster Trust in a Customer Dialogue System?

Abstract

In this poster paper we present a web user study about a customer dialogue system, in which participants assigned tickets based on an automatic classification to different departments and answered questions about the perceived classification performance. Completion times were significantly shorter when offering explanations on the classification process, while task success and trust in the interface did not depend on showing explanations or not. Based on the results, future studies should be confined to smaller scopes and investigate more techniques for explainable AI.