The double-edged sword of proactive GenAI: how it shapes user usage intention

Abstract

With the emergence of generative AI (GenAI), proactive services that initiate actions on behalf of users without explicit prompts are becoming increasingly prevalent. Although such services can enhance human-AI interaction, emerging evidence suggests that GenAI proactivity may also elicit adverse user reactions. To reconcile these mixed findings, this research examines why and when GenAI's proactive services positively and negatively influence users' usage intention. Drawing on mind perception theory, we develop a dual-pathway model proposing that increased proactivity simultaneously enhances perceived agency and diminishes perceived control. We further propose that task type (hedonic vs. utilitarian) determines the relative dominance of these two pathways. We test these predictions through two scenario-based experiments. Study 1 examines the dual-pathway model in a health assistant context, whereas Study 2 investigates the moderating role of task type in a shopping assistant context. Study 1 demonstrates that high (vs. low) proactivity simultaneously increases perceived agency and reduces perceived control. Study 2 shows that task type moderates these effects: for the hedonic task, the positive pathway through perceived agency dominates, making high proactivity more favorable; for the utilitarian task, the negative pathway through perceived control prevails, making low proactivity more preferred. This research advances the literature on GenAI's proactive services by demonstrating that their double-edged effects operate through parallel enabling and constraining pathways. We also extend mind perception theory by revealing the psychological costs of agency attribution. Practically, the findings offer actionable insights for firms seeking to tailor GenAI functionalities to enhance user experience and foster usage across diverse digital business contexts.