That is a good point, but as I mentioned, this is only the start of the conversation. Implicit bias is an easy thing to measure, but it has predictive issues in non-test situations. However, as it is always in science, it is the flawed experiment that led to better experiments (I am picking up the "popsci" stars, for this conversation to be rigorous we have to include actual meta studies, but you will have to contact the professionals for that):
- the general idea of "stereotype threat" (which becomes unrelated, not as much of an offshoot)
I am not expecting you to spend the time to vet every single of those links (and admittedly I used google, so some of the links might be overly editorialized), however I do believe these are good resources to consider for inclusion in your intellectual toolkit when you have the time.
This was a discussion a month ago. I'm mostly basing my opinion on it.