Identifying and Addressing Ethical Challenges in Recommender Systems

Simultaneously with the internet and the artificial intelligence technologies evolution, recommender systems have also evolved and their use has been established across the spectrum of digital life. All content, (information, people, objects etc.) which we process, buy or study, reaches us through a recommender system filter. After all, internet browsing would be impossible without them, since information overload would paralyze any attempt. Nevertheless, many concerning issues identified over the last few years have - with more or less certainty - been connected with the recommender systems activity. Issues of privacy, personal data, fairness and transparency, but also issues concerning personal identity and the proper functioning of society and democracy have risen. In the publication at hand, the recommender systems technology is analyzed with a focus on the different techniques and effectiveness measures employed to address the aforementioned issues. Furthermore, different problematic sides of recommender systems operations are identified and categorized, and a measure of a recommender system’s criticality is developed. Different case studies are presented, under the perspective of criticality, based on the proposed measure. Finally, a series of regulatory actions is proposed. The goal of the latter is the betterment of the systems’ function and the harmonization of their progress with social prosperity.