Redlining laws are outdated. If an ML algorithm selects zip code as a predictor of credit risk, it's because it is, not because the algorithm is using it as a cover for racism. Duh!
-
Of course. Imaging having rich parents was correlated w being good at statistics and we have no real stats features. But imagine you are a great statistician w poor parents. Algorithm chooses someone worse than you because they got rich parents. Of course that is unfair no?
There's nothing unfair about false negatives (or positives). Get better features, or don't be surprised if the results are poor.
-
-
The results are bad. We agree. But they are also unfair. Or don’t you agree?
-
And maybe to clarify the definition of unfairness I use. A feature of a person that they can not change and that is not causally related to their performance is used to make a decision against them.
Keskustelun loppu
Uusi keskustelu -
Lataaminen näyttää kestävän hetken.
Twitter saattaa olla ruuhkautunut tai ongelma on muuten hetkellinen. Yritä uudelleen tai käy Twitterin tilasivulla saadaksesi lisätietoja.