3 ms·
The language used in the article seems political/inflammatory to me. It's not a lack of "fairness", it's just a lack of accuracy. Imagine that you train a mod
by throwaway22032 2y ago
The language used in the article seems political/inflammatory to me.
It's not a lack of "fairness", it's just a lack of accuracy.
Imagine that you train a model to find roofing issues or subsidence or something from aerial imagery. Maybe it performs better on Victorian terraces, because there are lots of those in the UK.
Would you call it unfair because it doesn't do so well on thatched roof properties? No, it's just inaccurate, calling it unfair is a value judgement.
Bias is better because it at least has a statistical basis but fairness is, well.. inaccurate...
- CoastalCoder 2y agoThis was my general take as well. Although now that I think of it, there could be a "fairness" angle regarding how many training data are obtained from each population group. But if that's what the author meant by fairness, it wasn't clear.
- KingOfCoders 2y agoI think how a scientist would use 'fairness' - e.g. mathematicians - and the general public would use 'fairness' are two different things. On top of that fairness in the general public is undefined as well. I once wrote on the many facets just concerning fair salaries here: https://www.amazingcto.com/fair-remote-developer-salaries/ https://www.amazingcto.com/fair-remote-developer-salaries/