We all have biases that we should try to be aware of. Our implicit prejudices may be at odds with our explicit attitudes. One problem when discussing issues such as racism and sexism especially is that surprisingly many people seem to think that such things have been largely dealt with in the 20th Century and are now of minimal importance.
https://implicit.harvard.edu has several tests designed to measure our implicit biases. As with any scientific test, there could be issues with methodology etc and, in addition to discussion of implicit biases (e.g. the psychology of them, how they affect our skepticism), that also seems an appropriate topic for discussion here.
Perhaps if I approach it from another direction. Taking the test, I was reminded of Gregory’s rather explicit assumptions about atheists. But not in a good way. I felt like Gregory was the test designer, and that biased assumptions were being made about me.
I do not like thought police. I accept the need to root out explicit bias in public policy and in the workplace. I also accept discussion of implicit bias among friends. But this test did not strike me as friendly.
I felt the same way. It seemed condescending as well as result-oriented.
petrushka,
Nor do I!
I think that’s a mistake. The experimenters aren’t saying that you shouldn’t have implicit biases or that you’re a bad person if you do. We all have them, and we can’t help it.
The experimenters know that, but the public generally doesn’t, which is why they strongly advise against using the IAT for anything other than purely educational purposes.
I think some here misunderstand how the IAT test works. This may be a function of which test(s) they have tried.
Walto – there is no effort in the test design to instill a bias, nor is there any measurement of inconsistent answers.
Yes, the test is frustrating to take, and there are ways to ‘game’ the test. There are also algorithms that attempt to detect such ‘gaming’. It’s not a perfect test, but it is a rather interesting way of accessing your “System 1”, which is otherwise hidden from your conscious thought processes.
The test works by measuring how long it takes you to respond. Response time is shorter if the attributes that are paired on the same side are attributes that your system 1 associates with each other. It’s an extension of the famous Stroop test, which is extremely frustrating…
I took the Gender vs Science/Liberal arts test, with the result that I have a ‘moderate’ implicit bias. This, despite the fact that my grandmother and aunt were doctors, and my mother was a chemist. Both my current wife and ex-wife are biologists. My second daughter is headed towards engineering, and her older sister is a statistician, ferchrissakes. I have (somewhat aggressively) pushed my daughters to pursue STEM careers.
The test is measuring implicit bias.
OTOH, the men in my family are engineers, doctors and a chemist. There’s a handful of lawyers too, but we don’t like to talk about that…
😉
There is absolutely an attemopt to instill a bias in the nif/laap test. That”s the whole point of it. I think you need to take that test before commenting on its virtues. And my criticisms of it have been muted. I think it’s comically bad. Confused, stupid even.
DNA_Jock, one thing I’m curious about in your post is youe claim that “…nor is there any measurement of inconsistent answers.”
That’s hard for me to believe with respect to the test I took. Many of the question in the early part were timed (and there was no instruction to hurry) but rather were repetitious questions about whether I agreed (moderately, slightly, etc) with such things as that people have the traits they have or biology factors into personality. These questions several times in slightly different ways. (Was there no such stuff in your test?]
Only later in my test (the matching portions) did I get the hurry instruction–along with a warning that if I took a long time, my results would be thrown out. So, I’m wondering how you know they don’t make questionable inferences with the early portions of the test. I mean, what else can one do with results like this: “He once said he slightly agreed with P, but he twice said he moderately agreed with P and he once said he only moderately agreed with -P.” ??
Again, I’m not suggesting that they try to instill a bias in the gender or race or etc. tests. But the Laap/Nif test is precisely an exercise in trying to get one to like Laaps more than Nifs and then see if they have succeeded. I really wish somebody else here could take that one so they’d see what I’m talking about. I received no response to my complaints about this test from the contact person.
Good point, walto, I was making some unwarranted assumptions here. I will try to give the Laap/Nifs test a go later today. I suspect that the Laap/Nifs test might be a sort of meta-test, to see if they can reproduce the effects in a totally manufactured challenge, or to explore the effectiveness of different ways of instilling bias.
I will note that it is fairly standard to ask almost equivalent questions multiple times, either to get a more graded response score, or to test the effects of “framing”, i.e. context. Testing for outright inconsistency is used in some on-line questionnaires in order to detect (and exclude) respondents who are “phoning it in” and providing random responses.
For example, I pay the (invited) respondents who answer my web-based surveys: if a respondent is providing inconsistent responses, we exclude the data and if the inconsistencies are blatant, then he/she doesn’t get paid. A secondary ‘tell’ is that these folks move through the survey faster than they could possibly read the rubric. Occasionally one will complain about not getting paid. That’s a pretty entertaining conversation. 🙂
walto,
I cannot find the Laap/Nilf test. What is its title?
From your description, I suspect that it is attempting to explore the “halo effect”.
At the end of the test, did it report to you some description of your implicit bias, as in “You have a moderate implicit bias against peopel with a dark skin tone?” ?
Thanks for your comments, DNA. I don’t know the name of the test. I just clicked on the link in the top post here, registered for a test, and that’s the one it gave me. The results said something like that I have moderate implicit bias and that I should ask myself a number of deep questions about my tendencies toward prejudice, including why assholes like me have been allowed by society to remain alive.
W
I looked at the test page shortly after this thread was started.
I noticed the need to register. And I decided that it was not worth the trouble.
The comments I’m seeing here are interesting. But they make me happy that I decided not to take the test.
I suspect many respondents were offended by the results they received. I know I was dismayed/disappointed. I will try to explain why I was not offended.
IMHO a lot of the firestorm produced by the Milgram experiment was fueled by the mistaken assumption that the poor subjects had been exposed to an awful truth about themselves that did not apply to regular people, such as myself (the Milgram critic). The IAT is subject to a similar, and I believe similarly misguided, criticism. (As noted above, the IAT is subject to other criticisms regarding what is it actually measuring. That’s another topic.)
For anyone who has read any experimental research on human psychology, for instance studies of visual or auditory perception, it should NOT come as a surprise that one’s brain is playing tricks on one. So, to be told “You have a moderate implicit bias” should not be deeply offensive. But it often is.
As a potential balm, I strongly recommend “Thinking, Fast and Slow” by Daniel Kahneman (Nobel in Economics). It offers a truly fascinating window into how our mind leads us astray, and also an introduction to what the IAT is actually attempting to explore – System 1 – the associations that we make automatically, unconsciously.
Neil Rickert,
No need to register, just use the “Or, continue as a guest by selecting from our available language/nation demonstration sites:” option.
You will be asked demographic data, including your zip code.
I want to emphasize that I have no illusions that I’m beyond racism/sexism/speciesism/etcism and that I understand that some of my biases may be unconscious (or as they call it, “implicit.”) I figure I’m as poorly arranged as the next guy/woman. But that doesn’t mean Harvard’s test(s) for my undoubted flaws are any good.
Keith linked to some critical remarks about this study. My guess is that those criticisms are on target (and that there are additional defects as well).
I took the Presidents and Asian tests, and got unsurprising results. Little to no implicit preference between Barack Obama and Ronald Reagan, but a moderate tendency to associate ‘European’ faces (vs ‘Asian’ faces) with America.
I’m not saying there aren’t flaws with the tests, but they seemed reasonable to me.
I agree. But you appear to be assuming that the test is useless, based on the facts that it was annoying and frustrating, and you didn’t like the results. You made this assumption with an incomplete understanding of how the test works.
The article points out that the IAT is measuring in-group vs out-group rather than specifically race. That makes sense to me. And it points out that implicit (or perhaps instinctive) bias is only a weak predictor of overt, functional bias. As I understand it, that’s because system 2 – your rational self – can override system 1’s unconscious associations if you make the effort. I repeat my recommendation of Kahneman’s book.
However, I hope you can see the irony of “My guess is that those criticisms are on target (and that there are additional defects as well).” in this context. You are demonstrating the “halo effect”.
As did I, when I first responded to you. 🙁
ETA: “attribute substitution” may be a more accurate term here. They’re related, of course.
socle:
My impression is that the implicit associations are real, and that the tests detect them accurately. How else could we explain the consistent differences in response times?
I think the results are real, but the interpretation of the results can be problematic.
Associations aren’t necessarily biases. More men than women were breadwinners when I was young, so I probably implicitly associate men more with careers and women more with family. That doesn’t imply that I implicitly believe that women should stay home while the men work; it could go either way. It just means that I’ve learned a particular implicit association through experience.
On the other hand, if someone implicitly associates homosexuality with laziness, or something like that, then that strongly suggests a real bias, not just an association. Unless that person happened to be raised among a large group of lazy gays. 🙂
I need somebody (ANYBODY) to take that laap/nif test! I keep being put in a position of seeming to attack the other tests when I’m just criticizing THAT one.
walto,
I looked for that test but I can’t find it. Do you remember exactly where it was? Maybe you can look it up in your browser history.
I found a complete description of the test, its methodology, and their results here:
http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&ved=0CDQQFjAB&url=http%3A%2F%2Fapollo.psico.unimib.it%2Fshared%2Fpsychoscope%2Fsite%2Freference_pap%2FRN_JESP2010.pdf&ei=BTtAU9biBMbmsATy7IGgAg&usg=AFQjCNGY8rkOSoc1nkXntJV3EpbW1w0A2Q&sig2=6qR4o3BrqYzHPX9JiC6aDQ&bvm=bv.64125504,bs.1,d.cWc
I especially direct attention to the discussion in the last few paragraphs prior to the acknowledgements. That’s where most of the handwaving occurs regarding why the investigators’ assumptions were correct in spite of their results being contrary to their expectations. I think in fact, nothing could have changed their minds (which is always a nice way for scientists to be).
Walto, thank you very much for finding this. If, as seems likely, the test you took was the same as Study 2 in this paper, I fully sympathize that it must have been extremely annoying to take. You have my sincere sympathy.
But I am still puzzled: this paradigm is designed to distinguish the roles of the implicit vs explicit attitudes in the formation of the illusory correlation. As such, I fail to see how it could possibly be used to provide you with much in the way of relevant personal feedback. The most it could manage is “you have a {strong|moderate|weak} implicit tendency to favor in-groups” or some such bromide. Sounds like they are over-concluding at Harvard.
To repeat, you have my sympathy. But your final paragraph is wrong, and somewhat rude. Which makes me believe that you would be less likely to help an old lady across the road. That’s just the whacky way my brain works.
And yours, too 😉
Thanks for your sympathetic words, DNA. I agree with you that they are over-concluding at Harvard.
Can you explain why you think my last paragraph is wrong, though (I mean leaving aside my final swipe)? Did you see they way the paper hems and haws at the end about the apparently unexpected result that the implicit reaction conforms more with what they take to be correct than the explicit reaction? (I admit, however, that it’s pretty hard to follow, so I could be mistaken.)
walto,
Their data support their stated, if almost incomprehensible, thesis, i.e. “The distinctive influence of these two processes may lead to dissociation between implicit and explicit attitudes when reasoning beyond activated associations influences explicit evaluations”.
Or in English : “the explicit responses will reveal a manufactured correlation, but the implicit attitudes will not make this mistake” or “sometimes, your intuition is better than your analysis”
When they ask “Why then did this not occur?” the word “this” refers to what might be expected if Fiedler were correct, not their own prediction. I’ll grant you that the paragraph that begins “While our results…” contains some fairly impressive hand-waving, but it is of the ‘Fiedler might be yet salvaged’ variety.
I agree. I wonder if they put any ‘unexpected’ associations in the test as controls, like lazy ~ gay. As fair as I can tell, the tests all relate to associations which we expect some people to make.