With ‘AI-human teaming’ on the rise, different perspectives exist on how these novel forms of collaboration link to human wellbeing. Our goal was to understand how humans perceive collaborations with an AI agent from a ‘team process’ lens. Building on job demand-resources theory, we propose that individual perceptions of human-AI collaboration processes are negatively associated with human strain because AI helps people achieve goals and reduce their job demands. More importantly, we propose that AI agent level of control is a critical moderator for the link between individual perceptions of human-AI collaboration processes and human strain. Drawing on over 490,000 tweets about #ChatGPT, we analyze tweets specifically focusing on human-AI team collaboration. Using a mixed-methods approach combined with computer-aided text analyses (CATA), we explore how humans engage in (or anticipate) interactions with AI agents falling into transition, action, and interpersonal processes. Furthermore, we index AI agent control levels and human strain using CATA. Our results show that individual perceptions of human-AI team processes are negatively associated with human strain. We also find support for AI control moderating this relationship, that is, when AI agent control is high, human-AI action processes are positively associated with human strain.
