You had me and then you lost me. This quote betrays some naïveté about how learning (either in the Bayesian or behavioral sense) works. Depending on the nature of the belief we’re talking about, if your priors are solidly formed, a single anomaly isn’t going to be sufficient to overturn the belief. It takes some reinforcement.

Also: sometimes those priors are formed on actual experience, not just free-floating stereotypes.
Also also: sometimes prior beliefs, justified or not, can act as self-fulfilling prophecies, in that they can bias interactions toward reinforcing priors rather than undermining them.

It’s interesting to think about though. I wonder if anyone’s done any serious work on this. My intuition is that it would not be hard to train a computer to be “racist” based on some rules most people would agree to and real-life data, with obvious implications for many of the charges of racism that get thrown around so casually these days.

I don’t think the point was that if the person in question acts contrary to your established stereotype, you should re-evaluate that stereotype. Rather, you should stop trying to apply the stereotype to that person, and recognise them as the individual they are.

“Unfortunately a lot of people have a hard time updating their [beliefs] based on new information, which is what you should be doing (quickly) when you learn more about a specific individual. (emphasis added)”

And yet, see table 10.3 in “The Unbearable Accuracy of Stereotypes” and read the discussion You will see that the statement above is meta-analytically unsubstantiated. In the real world, People discriminate primarily on the basis on individual, not group characteristics.

Jayman, I am not sure about this. I used to talk with my daughter about the HBD, and I always tried to tell her that she should always differentiate between an individual and the population means. However, quite recently during a discussion about one of her collegues in the class it came that the collegue is a person of partial-black ancestry, and she said several openly racist statements. This disturbed me quite greatly. I mean, if I spent hours to explain what a mean is, a median, an average, to explain that she can’t judge people just by the average of the population they came from – and she still cames with abrupt, racist statements – damn, what would be reaction of the population if the HBD concepts would start to be popularised by the media?

I still have to find out a method of popularisation of the knowledge, which would not lead to the spread of racism at the same time.