Learning the algorithm: Responsibilization and the limits of agency on YouTube

Content creators invest extensive effort in “learning the algorithm” to secure visibility and growth. While existing research documents the proliferation of folk theories and optimization strategies, less attention has been paid to the structural conditions under which such learning unfolds. Drawing on a computational analysis of 179 YouTube videos discussing the platform’s algorithm, this article identifies recurring interpretive patterns that frame visibility as optimizable through metrics, strategic adjustment, and disciplined self-management. This paper argues that attempts at algorithmic learning are sustained by an asymmetry between epistemic agency – the capacity to interpret metrics and hypothesize about system behavior, and effective agency – the capacity to reliably influence outcomes. Because large-scale machine-learning systems are probabilistic, multi-objective, and continuously recalibrated, effective control remains systemically indeterminate. This asymmetry produces governance through responsibilization: creators internalize responsibility for unstable outcomes, sustaining engagement because the system appears responsive while remaining structurally uncontrollable.