Abstract

The growing use of conversational generative AI agents by an increasingly diverse set of individuals gives rise to the following question: How do individuals calibrate their trust in a dynamic, seemingly general-purpose technology that learns from and adapts to their behavior and may take an increasingly central role in their online interactions? In this article, we propose the first steps towards an integrative framework for trust calibration in genAI agents focusing on individuals using these agents in their daily lives. We address the notion of calibrating trust in technology, reviewing five challenges to our understanding of trust calibration in such context, namely (1) the complex nature of individual private use, (2) the complexity in assessing actual trustworthiness, (3) the fuzzy notion of the trustee, (4) the complex ontological status of these agents, and (5) the over-time nature of trust calibration. We conclude with a set of propositions for trust calibration in genAI agents and considerations for future research.