Reverse Engineering of Content Delivery Algorithms for Social Media Systems

Social media systems have become a primary platform to consume and exchange information nowadays. As a black box social algorithms were designed and trained to pick up filter and rank the most relevant and desired content to be delivered to each individual one of us. However these modern social algorithms were typically complicated and its development involved easily more than one hundred software engineers. Therefore we including the social media providers do not really know how these social algorithms work such that we are unsure about the quality of the delivered content or the existence of any bias or disparity? In this paper we explore content delivery algorithms on Facebook fan pages and communities. We empirically defined four hypotheses which originally were from Facebook features social network concepts and what we have observed from our SINCERE data set. We took a heuristic approach to statistically analyze our data set to uncover probabilistically how Facebook push user-created contents among users. Each hypothesis is represented by an explainable logical expression rule. We used each of the top 100 posts with the largest number of comments in ABC News CNN and Fox News fan pages. Our main contribution is to validate each hypothesis against empirical data such that together we provided a partial explanation for the content delivery algorithm used by Facebook.