hashtagFairness is hard, hashtagAI or not

👉 Every time I think of ways to make decisions fairer, I realize very quickly that fairness is not a one-dimensional concept, in fact it is a high-dimensional, multi-faceted ball of complexity. Not only may fairness apply to inputs or outputs (with very different consequences), it also applies to ANY feature of relevance (gender, race, age, socio-economic status, and many more), leading to a multi-criterion decision problem. Anywhere along the Pareto hypersurface could be deemed "fair" but, again with very different consequences.

💡 And then I remember this fantastic 2019 MIT Technology Review article Can you make AI fairer than a judge? Play our courtroom algorithm game (link) that I recommend to anyone who wants to EXPERIENCE how hard it is to consider all variables when trying to make fair decisions. The decision maker has to make a personal judgment call and pick a tradeoff that is acceptable to them.