3 ms·
When I was reading through some algorithmic fairness literature some time ago, I came back a bit frustrated because as the article mentions, the fairness defini
by Sol- 8y ago
When I was reading through some algorithmic fairness literature some time ago, I came back a bit frustrated because as the article mentions, the fairness definitions are mutually incompatible (though some seem more plausible than others) and it's not really a problem that can be fully solved on a technical level. The only flicker of hope was that a perfect classifier can, by some definitions, be considered fair, so at least you have something to work with - if your classifier discriminates by gender or other attributes, you should at least make it good enough to back up its bias by delivering perfect accuracy (at which point you can investigate why inherit differences between groups seem to exist).
It's good that some Computer Science researchers are ready to work in such politicized fields though, it's definitely necessary. I find it admirable because I personally wouldn't enjoy those discussions.
- LoSboccacc 8y agoIf unconstrained learning emits biased result it was given biased samples. That or the bias is in the dataset itself, which could still be fixed by removing and randomizing traits but at which point your alghorithm is learning a representation if reality which usefulness depend on the realm if application, say, great for university admittance and not so great for medical insurance purposes.
- tomjen3 8y agoYou assume that the underlying data cannot be correct and not fair.
- LoSboccacc 8y agouh, no? > [then] it was given biased samples
- commandlinefan 8y agoWell, whether or not (and if so, to what extent) unconscious bias taints engineering or scientific research is an interesting, and potentially important question to answer, it seems to me that the only people who are addressing the question are people who are overtly, consciously biased themselves - to the point where they actively exclude anybody who doesn't share their own set of biases.