2 ms·
What would it take for you to falsify your stance? If nothing can make you falsify it, then that's not science, that's pseudoscience. Unconscious bias entering
by xamuel 7y ago
What would it take for you to falsify your stance? If nothing can make you falsify it, then that's not science, that's pseudoscience.
Unconscious bias entering AI is bad. But consider this scenario. Imagine for decades, there's been a chemical leak in Cleveland, which has been causing people there to become worse at math. An AI detects that people from Cleveland are worse at math. Do you just throw the data away because it must be subconscious human bias? If you do, you're throwing away a chance to discover the underlying cause (the leak)! Who knows what sort of things we've thrown away on the altar of "everyone must be equal, there can be no nuance"!
- kaitai 7y agoThis is a straw man argument. It's easy to find examples of algorithms that are biased -- let's just look at facial recognition. Sure, there are situations like what you describe, but are you going to argue then that dark-skinned women really must be men, or more manly, or something, just because Amazon's facial recognition algorithm has trouble with dark-skinned women? The parent is not making a "scientific" claim. They are making a workflow/process claim. It is not the case that all things are "science" or "pseudoscience". There are many other topics humans can discuss.
- xamuel 7y agoI'm not saying no biased algorithms exist. I'm just saying, we shouldn't throw away what initially appears to be discriminatory bias, we should instead ask, "what caused this?" And that goes no matter what color of skin is represented on the dev team.