don't get dragged into that false dichotomy. treating fairness as a merely technical problem is a cop out. in addition to looking at dataset and algo, look at the processes that generate data, how the data is collected/selected for the data set, and how the algo is deployed and to what end.