Oh, another thought. (sorry for taking up so much space!) Sometimes something looks really icky, such as evaluating a candidate via religion, but is actually just standing in for a different trait. We care about A, and B is somewhat predictive of A, and A is really hard to measure, then maybe people sometimes use B as a rough proxy for A.
I think that this is sometimes used as the justification for sexism/racism/etc, where the old-school racist might say “I want a worker who is A, and B people are generally not A.” If the relationship between A and B is non-existent or fairly weak, then we would call this person out for discriminating unfairly. But now I’m starting to think of what we should do if there really is a correlation between A and B (such as sex and physical strength). That is what tends to happen if a candidate is asked to do an assessment that seems to have nothing to do with the job, such as clicking on animations of colored balloons: it appears to have nothing to do with the job, but it actually measures X, which is correlated with Y, which predicts on-the-job success.
I’d rather be evaluated as an individual than as a member of a group, and I suspect that in-group variation is greater than between-group variation, echoing what you wrote about the priors being weak.
Oh, another thought. (sorry for taking up so much space!) Sometimes something looks really icky, such as evaluating a candidate via religion, but is actually just standing in for a different trait. We care about A, and B is somewhat predictive of A, and A is really hard to measure, then maybe people sometimes use B as a rough proxy for A.
I think that this is sometimes used as the justification for sexism/racism/etc, where the old-school racist might say “I want a worker who is A, and B people are generally not A.” If the relationship between A and B is non-existent or fairly weak, then we would call this person out for discriminating unfairly. But now I’m starting to think of what we should do if there really is a correlation between A and B (such as sex and physical strength). That is what tends to happen if a candidate is asked to do an assessment that seems to have nothing to do with the job, such as clicking on animations of colored balloons: it appears to have nothing to do with the job, but it actually measures X, which is correlated with Y, which predicts on-the-job success.
I’d rather be evaluated as an individual than as a member of a group, and I suspect that in-group variation is greater than between-group variation, echoing what you wrote about the priors being weak.
You don’t need to apologise for taking up space! It’s a short form, write what you like.