Large technology companies hold unprecedented political, economic and social powers, given the ubiquity of recommendation systems in social networks, video and music streaming services, e-commerce sites and many other services. These systems select metrics such as engagement and clicks, which end up inadvertently amplifying sensationalist and misleading content, compromising society's access to reliable sources, information and issues of public interest.
The main platforms today are based on algorithms that perform an editorial function guided by criteria defined and programmed by humans, as they select content that will be exposed to users. Given the lack of transparency, the political and economic interests of companies are covered by the technical veneer of the algorithms they use. On the other hand, content producers constantly seek to adapt their production and distribution practices to the opaque rules of these systems, at the risk of having their reach and financing harmed by the platform if they do not adapt.
Furthermore, recommendation systems are characterized by microsegmentation, using data produced by users themselves as input for the accuracy and relevance of recommendations. Given the opacity in many of the methods used in the areas of data mining, big data and artificial intelligence, different studies have pointed to the incidence of worrying biases reinforced by these algorithms. Despite a friendly and conciliatory speech, technology companies have been accused of hindering research and data auditability, preventing a better understanding of the biases of recommendation systems. Academic literature has been criticizing the self-regulation and transparency initiatives of online platforms, pointing out the need for audits that would shed light on the internal functioning of the algorithms used.