[2209.05550] Mathematical Framework for On-line Social Media Auditing

Obtain a PDF of the paper titled Mathematical Framework for On-line Social Media Auditing, by Wasim Huleihel and Yehonathan Refael

Obtain PDF
HTML (experimental)

Summary:Social media platforms (SMPs) leverage algorithmic filtering (AF) as a method of choosing the content material that constitutes a person’s feed with the goal of maximizing their rewards. Selectively selecting the contents to be proven on the person’s feed might yield a sure extent of affect, both minor or main, on the person’s decision-making, in comparison with what it could have been beneath a pure/truthful content material choice. As we now have witnessed over the previous decade, algorithmic filtering may cause detrimental unintended effects, starting from biasing particular person choices to shaping these of society as an entire, for instance, diverting customers’ consideration from whether or not to get the COVID-19 vaccine or inducing the general public to decide on a presidential candidate. The federal government’s fixed makes an attempt to control the hostile results of AF are sometimes sophisticated, as a result of paperwork, authorized affairs, and monetary concerns. Then again SMPs search to observe their very own algorithmic actions to keep away from being fined for exceeding the allowable threshold. On this paper, we mathematically formalize this framework and put it to use to assemble a data-driven statistical auditing process to control AF from deflecting customers’ beliefs over time, together with pattern complexity ensures. This state-of-the-art algorithm can be utilized both by authorities performing as exterior regulators or by SMPs for self-auditing.