The Web Cannot Guarantee Content Diversity by Design: The Case of A/B Testing

The Web has often been portrayed as the most promising platform for the free dissemination of content of all kinds. In this vision, heterogeneity and diversity in content distribution are seen as essential guarantees of a democratic digital environment. However, in practice, the Web is largely managed by private corporations that not only provide access and services but also actively shape and adapt content to individual users. This dynamic is most prominently observed within social networks. Given this hierarchical and profit-driven structure, a critical question emerges: Can the Web, in principle, serve as a fair and open framework for the free dissemination of content? To explore this question, I construct a synthetic framework—an agent-based model—that simulates a social network structure. This model is designed to assess the impact of algorithmic mechanisms used by web companies on collective user behavior. I focus specifically on a widely employed strategy in the digital economy: A/B testing, a method used to enhance user engagement and optimize content delivery. In this study, I present the effects of a prevalent content selection mechanism that challenges the ideal of a pluralistic and democratic World Wide Web. My findings indicate that A/B testing has a profound influence on the qualitative dynamics of information dissemination within social networks. By design, its implementation systematically undermines the persistence of content heterogeneity and diversity across the Web, inexorably favoring content polarization.