3 ms·

Fairness in statistic is not not a subjective concept, it's a property of an algorithm.

• Thiez 8y ago
  Fairness is absolutely a subjective concept. E.g. in the context of school choice, which is fair? Let's say there are 20 students applying to a school, and there are 10 spots available. a) All students applying to the school have the same chance to get a spot: 50% b) The best 10 students (by some definition of 'best', e.g. grades or an entry exam) get in. c) The 10 students with the richest parents get in. d) The 10 students that live closest to the school get in. We could go further: when 15 of the applicants are boys and 5 are girls (or vice versa), wouldn't it be more fair when 5 of each are chosen? How about racial distributions? Ultimately how you define fairness is subjective. Obviously once you have defined fairness you can attempt to measure it using mathematics / statistics.
  • zwily 8y ago
    The person you responded explicitly said “fairness in statistics”.
    • Thiez 8y ago
      What does "fairness in statistics" even mean? Perhaps there is a very obvious definition that I don't know about? Would you please define it for me, and (if this is non-obvious) explain how it relates to the school choice problem?
• gowld 8y ago
  Fairness is measured relative to a choice of equivalence classes and a choice of loss function. In all of optimization theory, the objective function is subjectively chosen.