Feels like the tricky part here isn’t computing a “fair” outcome, but defining what fairness even means in the first place.

Once you formalize preferences into something comparable, you’re already making a lot of assumptions about how people value outcomes.