I mention above, it's not the existence of the correlation that's a problem. Whether or not its "statistically significant", it's wrong to use correlations for deviation making on the individual level, because they are just stereotypes, they're not necessarily true. For population studies or something, sure, but same as you can't tell someone that just because someone with their personal attributes and history is statistically likely to be a criminal so they must be, you shouldn't be telling people they're dumb based on some abstract instrument just because its outcomes exhibit a correlation