Affect between Humans and Conversational Agents: A Review and Future Research Directions

Conversational agents (CAs) are digital artifacts which communicate with humans through natural language for performing a variety of tasks. When humans interact with CAs, affective responses e.g., emotions, are seen to influence both CA adoption and performance. Despite the vital role of affect in human-CA interaction, understanding of its antecedents, affective outcomes, and their relationships from prior research is nascent and segregated. Motivated thus, we review empirical studies on affect in human-CA interactions, classify them in terms of affective outcomes, and identify corresponding antecedents. We further analyze the relationships between antecedents and outcomes and highlight significant relationships as well as inconsistent findings. Drawing on the review, we propose future research directions for this area. We contribute by developing a deeper understanding of research on affect between human and CAs, identifying gaps in prior knowledge, and outlining future research directions. Additionally, we lay out our plans for extending this work