3 ms·
I get the positive feedback loop issue. I’m saying if I’m looking at the data (which is blind a -it doesn’t classify people by ethnicity) but manually when loo
by mc32 8y ago
I get the positive feedback loop issue. I’m saying if I’m looking at the data (which is blind a -it doesn’t classify people by ethnicity) but manually when looking at and labeling the data we find the above proportions as the effective proportions, then I think we’re okay.
- gbrown 8y ago> I’m saying if I’m looking at the data My point is about the data generating mechanism - making inferences conditional on a bad selection procedure is a bad idea. It would be bad statistics, and it's bad ML. In this case, my argument is that it's also bad for society.