Operational Mediation: Critically Theorizing Recommendation Systems

Abstract

Recommendation systems no longer merely suggest content—they define the architecture of digital mediation itself. From the divinatory algorithms of the I Ching to the black-box neural networks of TikTok, the evolution of RecSys reveals a shift from human-guided decision-making to autonomous, self-iterating computational epistemology. This thesis argues that RecSys have transcended their original function as assistive technologies, becoming omnipresent, unknowable, and posthuman infrastructures that shape visibility, knowledge, and agency. Through historical analysis, theoretical critique, and a prototype experiment, we examine how recommendation systems mediate digital culture—not by reflecting human preferences, but by recursively generating them.