The bias report everyone quotes has a math error on page 14

I was prepping for a debate on facial recognition and found the original study the whole hysteria points to. The false positive rate for darker skin was actually computed with a sample size of 23 people, which is statistically useless. Has anyone else actually pulled the raw data on these headline numbers, or are we all just citing the summary?

1 comments

Log in to join the discussion

Log In

1 Comment

ben_lewis11d ago

Wait, did you check if those 23 people were even representative of the broader darker skin group they claimed to test, or was it just one specific demographic? Because if the sample was mostly one ethnicity or age range, that error gets even messier than the math itself. I'm asking because the whole argument hinges on whether the follow-up studies used the same flawed data or actually re-ran the tests.

4