Fair Clustering

Abstract. Fair clustering is usually defined as clustering with each subgroup having approximately same representation in every cluster. This concept is important as given the blind application of the algorithm the biases of the data can be exposed. However, the biggest challenge of the topic is the ambiguity of the fair clustering definition and the notations the definition is based on. In the session "Fair Clustering" the reviewed papers Chierichetti et al. 2017, Bera et al. 2019, Huang et al. 2019 are featuring different algorithms providing solutions to fair variants of classical clustering problems, such as k-means and k-median problems. This survey paper focuses on the major ideas of the session "Fair Clustering" from the NeurIPS 2019 conference. We present main contributions of each paper and discuss the difficulties of defining the notion of fairness and fair clustering problem in general. Moreover, we conduct a systematic literature review of Chierichetti et al. 2017 in order to provide a solid foundation on the prior use of fair clustering. Then we elaborate on the comparison of the three papers and discuss the ethical moment of the notion of fair clustering as the conclusion of our work.