Clustering Algorithms

Clustering is one of the more famous algorithms out there for n data mining, data science. If your training data comes from a system that is systematically discriminatory, then your output model is going to inherit that. I guess fair clustering in the group level sense would look at trying to have some kind of minimum and maximum bounds on the proportion of each group members showing up in these clusters. It feels a little less connected for me than the loan application problem, where i see the tangible label, how does fairness play or, what's the harm that could happen here if we're not very careful in the process?