Exploring the antecedents and mitigators of user uncertainty in generative AI use: A principal-agent perspective

As generative AI (GAI) becomes increasingly intelligent and autonomous, users encounter new forms of uncertainty arising from the evolving user-GAI relationship. Understanding this transformation and how associated uncertainty can be effectively managed is critical for fostering trust and sustainable GAI usage. Grounded in Principal-Agent Theory (PAT), this study redefines the user-GAI relationships from a principal-agent perspective and develops a nomological framework to explain how key technical characteristics of GAI influence users' perceived uncertainty and continuance intention in general GAI use settings. Using a mixed-methods design, we first conducted an exploratory computational text analysis of large-scale user-generated online content to identify major sources of uncertainty and key GAI attributes, which informed the development of a nomological framework. We then validated and tested the framework through a quantitative survey. Our results show that users’ fear of opportunism and information privacy concerns significantly increase perceived uncertainty, thereby reducing continuance intention. Conversely, GAI accuracy, transparency, and empathy effectively mitigate uncertainty by alleviating these fears and concerns. This study contributes to the literature by investigating user uncertainty within a redefined principal-agent structure and integrating PAT with technological features to build a comprehensive nomological framework. Practically, we offer actionable guidance for designing human-centered GAI systems that emphasize accurate, transparent, and empathic interactions to reduce user uncertainty and support sustained engagement.