The ethics of recommender systems

Abstract

Recommender systems are all around us; they can be found in news applications, YouTube, Netflix, the healthcare industry, and e-commerce. These recommender systems are influencing our choices and the information that is presented to us. This makes it crucial to think about the ethical consequences of these recommendations and possible solutions to ethical issues. In this thesis, we have identified the main ethical challenges of recommender systems, and we looked at one specific, promising solution called the secondary ethical layer. The secondary ethical layer is a general ethical filter which filters out any unethical recommendations based on cultural and personal preferences while also taking into account all the different stakeholders on which recommendations can have an effect (such as the user, provider, system and society). We have found that this solution can solve some ethical issues, specifically with regards to inappropriate content, unfairness (biases) and issues for society. It does not solve problems such as the lack of opacity and some privacy issues within recommender systems. This thesis identifies different key elements of the ethical layer and creates the fundaments on which a practical solution can be built.