Regulating algorithmic filtering on social media
share
Through the algorithmic filtering (AF) of content, social media platforms (SMPs) have the ability to influence users' perceptions and behaviors. Attempts to regulate the negative side effects of AF are often difficult to pass or enforce due to critical social, legal, and financial considerations. In this work, we address this multifaceted problem by proposing a unifying framework that considers the key stakeholders of AF regulation (or self-regulation). We mathematically formalize this framework, using it to construct a data-driven, statistically sound regulatory procedure that satisfies several important criteria. First, by design, it moderates the effect of AF on user learning. Second, it has desirable properties of online governance, including being normative and user-driven. Third, by illustrating the regulatory procedure in linear dynamical systems, we prove that it can align social and financial interests, which are typically at odds. Specifically, we identify conditions under which the regulation imposes a low cost on the SMP's reward (e.g., profits) and incentivizes the SMP to increase content diversity. We provide illustrative simulations.
READ FULL TEXT
