The role of AI-based artifacts’ voice capabilities for agency attribution

The pervasiveness and increasing sophistication of artificial intelligence (AI)-based artifacts within private, organizational, and social realms are changing how humans interact with machines. Theorizing about the way that humans perceive AI-based artifacts is, for example, crucial to understanding why and to what extent humans deem these artifacts to be competent for decision-making but has traditionally taken a modality-agnostic view. In this paper, we theorize about a particular case of interaction, namely that of voice-based interaction with AI-based artifacts. We argue that the capabilities and perceived naturalness of such artifacts, fueled by continuous advances in natural language processing, induce users to deem an artifact as able to act autonomously in a goal-oriented manner. We show that there is a positive direct relationship between the voice capabilities of an artifact and users’ agency attribution, ultimately obscuring the artifact’s true nature and competencies. This relationship is further moderated by the artifact’s actual agency, uncertainty, and user characteristics.