sozobe wrote:Tell me more about the psychologists, Thomas.
I was very happy that in the most recent elections, the polling for black candidates matched up really closely with actual vote tallies. People seemed to be telling the truth about who they would and wouldn't vote for.
Oh, it's not about
specific black candidates. My pessimism is about general questions like voting for "a" black canditates.
The strongest influence on my new pessimism is a talk that Harvard's
Mahzarin Banaji gave at the "Beyond Belief" conference, which I heard via webcast. Unfortunately
her talk (Session 7 in the link) isn't captioned. But if the Deaf community operates something like a captioning service, I would bring this talk to their attention. It's very good. I originally opened the link to hear Richard Dawkins, but Banaji managed to outshine even him. And that's just about the highest compliment I can make about any conference speaker.
One way in which Banaji demonstrated biases in the audience was the following, four-step test. In the first step, she fired up PowerPoint on her laptop and made it throw male and female names at the audience. The audience was invidet to say "right" for males and "left" for females", as fast as they could. Their responses came pretty fast, say in 300 milliseconds.
In the second step, PowerPoint delivered "carreer" words (such as "salary", "promotion", "boss") and "housekeeping words" ("kitchen", "children", "garden"). Same test, about the same response time.
The third step made things more complicated by combining the first two. Now the audience should say "right" for "carreer or male name" "left for "house or female name", or stay silent if it was neither. Although the question was more complicated the response time was about the same, say 300 milliseconds.
The fourth step delivered the punch line. Now the audience was made to say "right" for "carreer or female", "left" for "household or male", or stay silent if the word belonged in neither category. Now, suddenly, the response time increased drastically -- say a second -- and the end of the test almost disappeared in increadulous laughter by the audience, which couldn't believe what a hard time they had putting "men's name" in the same category with housekeeping and "woman's name" in the same category as "carreer". Most people in the audience were middle-aged, liberal scientists who believed in equality. In polls, the would have answered the question "do you think it's it normal for women to make carreers these days?" with an resounding "yes!". But on a gut level, as the test demonstrated, they still found it difficult to actually, instinctively think this way.
Near the end of the talk, Banaji invited the audience to go to her website,
http://implicit.harvard.edu , where more implicit tests for bias await the curious surfers. I took a few of these tests myself, and to my frustration found out that I myself seem to be a lot more prejudiced than I'd like to think. I can warmly recommend visiting the site; it's an enlightening, if somewhat sobering experience. (I'd thought of starting a thread about it, but somehow didn't get to it.)
So that's why I have become more pessimistic about people's biases, and about the power of polls to fully bring them to light.