Conference Agenda

Social media recommender systems increasingly function as infrastructures of attention, shaping what users encounter, what appears salient, and what remains outside their field of view. Transparency has therefore become a central platform governance response, yet prevailing approaches often emphasize disclosure rather than whether affected audiences can understand, evaluate, and act on what is disclosed. This paper distinguishes disclosure-centered transparency from interpretability-centered transparency, and argues that transparency should be evaluated through demonstrated user understanding and actionability. Two exploratory pilot studies provide preliminary support for this distinction. Across the studies, outcomes moved selectively rather than uniformly. Greater awareness of platform harms did not translate straightforwardly into stronger action-oriented preferences, while exposure to simulated feeds produced modest movement in noticing feed characteristics but little broader change in understanding, response, or sustained practice. Greater social media feed content variety likewise did not consistently improve interpretability-related outcomes. Although varied social media feeds may reduce repetition or broaden exposure, these benefits are not equivalent to understanding how content was selected or what meaningful actions users can take. We therefore propose three governance reforms: standardized, contextual recommender system explanations; user controls whose effects can be observed and verified; and independent, recurring audits of explanation accuracy, comprehension, accessibility, actionability, and persistence. Together, these reforms can help shift transparency from formal availability toward demonstrable performance and offer a practical basis for strengthening US and international platform governance frameworks.