Brief summary ... there's a recent study (currently undergoing peer review) which spurs an interesting discussion. Researchers administered a small portion of the GRE verbal exam to a number of people. As has been well-documented elsewhere, African-Americans perform worse on such exams. But then they gave the same exam shortly after Obama accepted the Democratic Party's nomination for President ... and the performance gap disappeared.
The segment then goes on to look at some work that Claude Steele has done, both looking at race and gender, with very similar findings. (I won't spoil the story ... listen to the podcast for yourself.)
If this is indeed verifiable (and there is some substantial science to be done here), this is utterly fascinating. The importance of words and images, even when one is aware of their bias, appears to be substantial.