The Principal–Agent Problem in User–Algorithm Interactions

Algorithmic curation increasingly shapes media consumption, making user–algorithm interactions crucial for media use and effects. Many qualitative studies have characterized these interactions as agency negotiations, yet our understanding of how users negotiate agency with algorithms remains limited. To advance theorizing on user-algorithm interactions, we leverage Principal–Agent Theory—an economic perspective on negotiation scenarios—for interactive communication between users (principals) and algorithmic recommender systems (agents). We explicate the potential of Principal-Agent Theory for the case of short-video apps (e.g., TikTok). Using dyadic agent-based modeling, we examine two widely discussed phenomena: perceptions of responsive algorithms that ‘know’ users, and users’ attempts to curate content by ‘training’ recommenders. By combining a rational choice perspective with computational modeling, we offer a blueprint for critically examining the dynamics of agency in algorithmic media, with implications for further domains such as well-being, privacy, and media governance.