It's a common argument that AI-based evaluations and predictions hsould be fair. But what is fairness? This article looks at different ways of defining it. For example, fairness may or may not take into account the influence of being a member of specific groups - a race, perhaps, an occupation, maybe, or a school division. Arguably it's imporrible to satisfy both ways of being fair at the same time. "Pessimistically," writes Brian Hedden, "we might conclude that fairness dilemmas are all but inevitable; outside of marginal cases, we cannot help but be unfair or biased in some respect." But perhaps there's a way to "to take a second look and sort the genuine fairness conditions from the specious ones." Hedden writes, "I think expectational calibration within groups is plausibly necessary for fairness." Perhaps - but it's going to depend a lot on how these groups are formed, as Hedden shows. See also: impossibility theorem.
Today: 3 Total: 99 [Direct link] [Share]
