A Review of Emotions in Human-Conversational Agent Interaction

Abstract

Conversational agents (CA), powered by natural language processing, have become increasingly popular across multiple domains. However, these agents often fail to communicate effectively with users, leading to poor adoption and task outcomes. Emotions are a fundamental aspect of such interactions, influencing use and adoption of digital artifacts. Despite their salience, understanding of the role of emotions in human-CA collaboration remains fragmented. Motivated thus, we review empirical studies on emotions in human-CA interactions. We synthesize the findings from the reviewed studies in terms of antecedents, emotion-related outcomes, and their relationships in the form of a descriptive model. Based on the synthesis, we identify knowledge gaps and propose directions for future research. Our analysis provides insights into the role of emotions in human-CA interactions and contributes to research in this area.