Recommender systems have the potential of helping users in finding relevant items in the online environment, and in many ways, they impact which content we consume. Thus, how fair these systems are affects us. A common fairness issue in recommender systems is popularity bias. However, research on this issue has mostly been focusing on the algorithmic side, and the user perspective has been more or less neglected.

In this study, the goal was to understand whether there is a correlation between algorithmic and perceived fairness in the context of popularity bias, and the study was conducted in a music recommender setting. Three different algorithms were used in the study, each generating recommended playlists with varying levels of fairness in terms of recommending both popular and less popular music items. By comparing how fair users perceived the different recommended playlists to be with the algorithmic fairness of the playlists, conclusions could be drawn on the relationship between perceived and algorithmic fairness. Moreover, it was examined whether two different factors, namely familiarity and satisfaction, have an impact on perceived fairness.

An online survey was conducted, and it was concluded that there is no correlation between perceived and algorithmic fairness, as the participants could not notice any difference in fairness between the playlists. Familiarity was shown to only have an impact on perceived fairness in terms of one algorithm, while satisfaction was shown to have a significant impact on perceived fairness across all algorithms. The results indicate that fairness, in the context of popularity bias, may not be of high importance to users. As opposed to concentrating on how users perceive this type of fairness in recommender systems, it might be more important to focus on other stakeholders, such as the providers.