When AI-Based Agents Are Proactive: Implications for Competence and System Satisfaction in Human–AI Collaboration

Abstract As the capabilities of artificial intelligence (AI) technologies continue to improve, collaboration with AI-based agents enables users to be more efficient and productive. Not only has the quality of AI-based agents’ outcomes increased, but they can now help proactively, and even take over entire work tasks. However, users need to be satisfied with the system to remain motivated to collaborate and engage with AI-based agents. Drawing on self-determination theory, a vignette-based online experiment was conducted that revealed that proactive (vs. reactive) help from AI-based agents leads to a higher loss of users’ competence-based self-esteem and thus reduces users’ system satisfaction. This effect is moderated by the users’ knowledge of AI. Higher (vs. lower) levels of AI knowledge cause a greater loss of competence-based self-esteem through proactive (vs. reactive) help. The findings contribute to a better understanding of help from AI-based agents and provide important implications for managers and designers who seek to enhance human–AI collaboration.