In today’s multi-choice media landscape, algorithmic curation plays a central role in users’ digital media consumption. This has led to new, collaborative dynamics in content selection, where users negotiate agency with algorithmic recommender systems. An example is the widely observed phenomenon of users attempting to “train their algorithm” on algorithmic media platforms like TikTok. However, the motivations, mechanisms, behavioral patterns, and consequences of these user-algorithm interactions remain conceptually and empirically underspecified—despite their importance for understanding everyday algorithmic media use. Specifically, theoretical underpinnings and quantitative insights are still lacking on how algorithmic curation intersects with established media selection processes (e.g., mood management), how agentic or in-control users feel when curating their algorithmically pre-selected content, and how these dynamics influence usage patterns (e.g., session duration or disengagement). To address these research gaps, my dissertation aims to explore and theorize the patterns of user-algorithm interaction in algorithmically curated environments. In this talk, I will first provide a short overview of my research examining how algorithmic curation shapes everyday media selection and (dis)engagement. Second, I will delve deeper into an ongoing dissertation project that applies Principal-Agent Theory, an economic model explaining agency negotiations in dyadic, collaborative constellations (Eisenhardt, 1989), to user-algorithm interactions and tests it through computational simulation (i.e., dyadic agent-based modeling).

About the speaker

Alicia Ernst is a PhD candidate in the Computational Communication Science group at the Department of Communication, Johannes Gutenberg University of Mainz. Her PhD focuses on everyday media selection in algorithmic environments, whereas her broader research interests include the impact of digital media on entertainment experiences and well-being, intensive longitudinal/in-situ designs, and digital behavioral data approaches.