It’s not very good. It’s like baking with a malformed sheet of metal. It’s gonna leave a dent in the cake, and if you just fix the dent in the cake, you’ll keep getting dents in all other cakes. The dent in the cake is an error, but the bad shape of the metal is also an error. The metal is an underlying cause. I think since we can literally see the metal’s shape, it’s easier to spot.
How often do people think about fixing biases? Idk why I think that’s a CF exclusive thing. It’s something that would look so complicated and boring to work on that no progress is made. Maybe people in fields they’re good at work on their biases somehow. They’re able to do a good job in spite of some biases. Idk
What if you had like 3 biases? Each bias makes 5 errors every hour or something. If you fix an error after an hour you’re going to have 4 other errors for a bias. I think you’ll have 14 errors to fix in total. You’re going to keep having errors to fix while the biases keep making them.
I don’t think you’re ever going to win cuz biases are going to keep creating errors. You’re only going to keep fixing those errors while the biases create more errors. It’s like cleaning up a spill caused by a leak and never fixing the leak.