Another classic study by Darley & Gross published in 1983, found that people applied a stereotype about social class when they saw a young girl taking a math test, but did not when they saw a young girl not taking a math test. Two attempts at exact replication have failed. And both replication attempts actually found the opposite pattern – that people apply stereotypes when they have no other information about a person, but switch them off when they do.

This is from Claire Lehmann, “How a rebellious scientist uncovered the surprising truth about stereotypes,” Quillette, December 4, 2015. The whole thing is worth reading.

Another excerpt:

Lest one thinks that these results paint a bleak picture of human nature, Jussim and his colleagues have also found that people tend to switch off some of their stereotypes – especially the descriptive ones – when they interact with individuals. It appears that descriptive stereotypes are a crutch to lean on when we have no other information about a person. When we gain additional insights into people, these stereotypes are no longer useful. And there is now a body of evidence to suggest that stereotypes are not as fixed, unchangeable and inflexible as they’ve historically been portrayed to be.