February 27, 2013
Peering into our blind spots
Katie Koch in the Harvard Gazette:
Mahzarin Banaji shouldn’t have been biased against women. A leading social psychologist — who rose from unlikely circumstances in her native India, where she once dreamed of becoming a secretary — she knew better than most that women were just as cut out for the working world as men.
Then Banaji sat down to take a test. Names of men and women and words associated with “career” and “family” flashed across the computer screen, one after the other. As she tried to sort the words into groups as instructed, she found that she was much faster and more accurate when asked to lump the male names with job-oriented words. It wasn’t what a pathbreaking female scientist would have expected, or hoped, to see.
“I thought to myself: Something is wrong with this damned test,” said Banaji, Harvard’s Richard Clarke Cabot Professor of Social Ethics, as she reflected during an interview in her William James Hall office on her first run-in with an Implicit Association Test (I.A.T.).
That Banaji specializes in creating just these kinds of assessments did nothing to change the results. But at least she can take comfort in knowing she’s not alone. In the past 15 years, more than 14 million such tests have been taken at Project Implicit, the website of Banaji and her longtime collaborator Anthony Greenwald.
What these curious test-takers, as well as Banaji and Greenwald, found was that many of us hold onto quite a bit of unconscious bias against all sorts of groups, no matter how unprejudiced we strive to be in our actions and conscious thoughts. It’s a counterintuitive, even unnerving proposition, and one that Banaji and Greenwald, a professor of psychology at the University of Washington, set out to explain for a lay audience in “Blindspot: Hidden Biases of Good People.”
More here.
Posted by S. Abbas Raza at 12:42 PM | Permalink
Comments
I hope people will go to the site and take the IAT. When I first took it, I did not like what I saw, but I could not argue with its fundamental accuracy -- it found "the unreconstructed me," not the person I have struggled to be.
Posted by: Elatia Harris | Feb 27, 2013 2:56:26 PM
Oh, the moralizing! Blind spots make us the contradictory beings we are. Who wants to belong to the "good people"?
Or the lovely German term "Gutmenschen".
I just remembered Nietzsche's "Morality is the herd-instinct in people".
Posted by: Orla Schantz | Feb 27, 2013 6:06:51 PM
Took several of these tests myself, found it fascinating and disturbing. I was not surprised that I had irrational biases but I was shocked by their extent.
The next time there was a race-component in the categorization task, I tried to counter-prime myself by thinking of specific black people I admire a lot before starting the exercise. This worked so well that the results even indicated a slight preference for black people over white people this time around.
Some degree of bias might be inescapable, but it's not as if awareness and effort could do nothing to lessen it. Hopefully, we can all agree that the effort is absolutely worth it.
Posted by: Adele Quested | Feb 28, 2013 9:08:31 AM
Is there a test that attempts to determine to what extent we are biased towards political correctness over our own experience?
Posted by: carlos | Feb 28, 2013 10:46:43 AM
You mean a preference for an attempt of fairness over proudly clinging to confirmation bias?
Posted by: Adele Quested | Feb 28, 2013 8:49:04 PM
Let me add that I come from a country that has a fairly low number of black people and I have probably not made much effort to meet any - on the top of my head, I can only think of a classmate's little half-sister, our parrish priest's colleague who stayed here for two years to complete his studies and various immigrants selling the local street magazine. So whatever perception I seem to have of black people is certainly not grounded in much personal experience. But if I'm drawing on media portrayals anyway why focuse on the negativ ones, as my subconscious seems apparently intent on doing?
Posted by: Adele Quested | Feb 28, 2013 9:10:28 PM
Adele -
The next time there was a race-component in the categorization task, I tried to counter-prime myself by thinking of specific black people I admire a lot before starting the exercise. This worked so well that the results even indicated a slight preference for black people over white people this time around.
But this basically makes the test rather less useful, istm. If you can get any result you want to by thinking suitable thoughts in advance (and I can confirm this having hit upon the same trick. It's even easier than thinking of positive role models actually. To get a result strongly in favor of one group, simply say mentally some strong expletive referring to the other group, and invest emotionally in that curse. Become temporarily bigoted in the other direction iow. It's a whole lot easier than holding only good thoughts about the group you want to make win) all the test is probing is what you're thinking when you attempted it.
In other words, when you take it _without_ having played mental tricks, the test measures in some sense a weighted average of messages and thoughts you've encountered about the test groups over the course of your life. NOT your moral attitudes about the groups. You "win" the test not by being morally enlightened or unbiased, but by holding no *beliefs* about groups that distinguish between them. To the extent that those beliefs are shaped by myths, errors and prejudice, the test helps, and you can use it as a personal diagnostic. But it's hard to think all such beliefs are held for those reasons. There's being a bigot, and there's being unlobotomized. IAT prefers people to have forgotten statistics, not to have given up on bigotry.
Posted by: prasad | Mar 3, 2013 12:27:36 PM
Good stuff Prasad!
And, what is the point of taking the test if you go other than as you are? Is not the point to test yourself and see how you can remedy what you find?
I recall my father was not very happy with my IQ test results. His own was 200 and mine shoulda been close, right? Not even. He told me to keep on taking the test until I figured it out better. The director of the testing service got rather tired of my doing this, although my score did improve a little. The director asked how I felt about it. "Eventually," I replied, "my father will figure out that I am not getting smarter, only testing better, and he'll drop it." Which is what indeed happened.
All this by way of posing a question: Do you want to use a test to change yourself, or do you want to head fake the test to score better? Developing the latter capacity is too often rewarded; genuine change in a direction you want to go, and the real learning that enables it, are experiential prizes best sought by the real you. And that journey starts in a spirit of truth-seeking, not test anxiety.
Posted by: Elatia Harris | Mar 3, 2013 1:09:01 PM
Elatia, I just watched a discussion at bloggingheads with Banaji in it, and left a long comment there. I'll paste this bit:
--
Banaji says she has a screensaver image of a female construction worker breastfeeding, to help counter her "biases." Notice that the test doesn't even pretend to probe attitudes, implicit or otherwise, about breastfeeding in the workplace or the appropriateness of people doing jobs that few others of their gender do. Instead you'll pass an IAT about jobs and gender if you've never observed that most people working such jobs are male, though apparently the screensaver trick works too. Being unobservant (or an idiot) shouldn't make it easier to pass tests, no?
--
In general I think a test like this mixes beliefs about the world with attitudes toward people too cavalierly. That has an ironic implication here - people who're trying to fix a problem will get the same bad IAT result as someone who's gloating over that problem's existence, as will anyone else who's simply aware of numerical trends. If the test tells you that you have a bias you have no conscious awareness of possessing, maybe you're right and the test is saying something else.
But suppose you've taken the test and you get a sense that you're biased. That happens too, and I can also say that the IAT has given me a better sense of mine. In such cases the proper use of the test is clearly to become more aware of your own biases and to try and counteract them. But I'm not sure the methods Banaji suggests (of exposing yourself to positive role models and counter-stereotypes) actually do this. In fact I think they're about gaming the test (by putting desirable ideas into your memory) and learning to ignore anything that's niggling. In this way the self-help technique might be rather like your example of taking tests over and over again. Which is not to say that the test is useless (any more than those other tests are) but that you need to be careful about using them in not letting the thing being tested come apart from the score.
So yeah, I think this test is hard to interpret, though useful as diagnostic. But the self-help technique proposed needs a lot of work. Better (though less sexy) seems the more obvious method - if you're on a hiring committee, know about your biases using among other things IATs as guides, and try to do your best. Even better, devise procedures like blinding where decisions are very open ended (say when hiring a chef)
Posted by: prasad | Mar 3, 2013 2:35:12 PM
Prasad, what you do about the results of the IAT depends on whether you think you are reprogrammable by means such as stereotype-busting screen savers. Behavioral psychologists have seen some good outcomes this way. But --are they real life outcomes or better test scores? Do you hire a nursing mother as the GC of your construction project? Or, do you reduce your biases in a way the test can discover, but that leaves your decision-making, when it comes to offering a job to a person you might once have pre-rejected, unchanged? This coming spring I will interview Mahzarin Banaji for 3QD and we will have some questions for her!!!
The world of classical music has taken measures that impress me to counter its hideous and acknowledged bias against women as orchestra musicians. In some auditions, the instrumentalist trying out is behind a screen. This has given a huge boost to women hires. Soloist auditions do not happen like this, however. The charisma of the solo instrumentalist is a big factor, and there are excellent of both genders musicians without any. But this is a good example of how hiring policies (practice) can be ahead of testing (theory).
Posted by: Elatia Harris | Mar 3, 2013 3:43:27 PM
I don't know much about Banaji's work, but I know enough about sexy theories promoted by social psychologists to know that most of it is rubbish. Here is a balanced review of the sorry state of the field: http://chronicle.com/article/Power-of-Suggestion/136907/
Most of these experiments are to put it mildly "non replicable". They merely serve as "scientific" cover for political agendas.
Posted by: Sundar | Mar 3, 2013 5:19:55 PM
Elatia Harris asks a revealing question: "All this by way of posing a question: Do you want to use a test to change yourself, or do you want to head fake the test to score better?"
And here in a nutshell we have the unstated goal of much work in social psychology: to prod people into "changing themselves".
Posted by: Sundar | Mar 3, 2013 5:26:42 PM
Elatia Harris writes: "Prasad, what you do about the results of the IAT depends on whether you think you are reprogrammable by means such as stereotype-busting screen savers. "
Elatia, your choice of words like "reprogrammable" quite revealing of how you view your fellow human beings. And this view is not new either. Numerous large scale "reprogramming" experiments have been tried in the 20th century.
Posted by: Sundar | Mar 3, 2013 5:45:40 PM
Sundar, I'm not sure I understand where you're getting a totalitarian vibe from our conversation, but it seems like self-improvement is some distance from the gulag.
Posted by: prasad | Mar 3, 2013 6:23:43 PM
Also, I don't know how Mahzarin Banaji views thing, but it seems very natural to view say racial prejudice as emerging directly from aspects of human nature, and it's easy to tell an evolutionary story about in-group preference or genetic relatedness as accounting for this. I'd be surprised to find Banaji endorsing a blank slate view of things, in other words.
Posted by: prasad | Mar 3, 2013 6:28:49 PM
Sundar, I think the IAT is a matchless way of finding out things about yourself you may not like. What does one do with that? Not the business of the IAT. Conceivably, a person who takes it will feel validated rather than exposed. Some people are very comfy in their racism and sexism -- they believe their point of view reflects the natural order of things.
Say -- will you help a girl out??? I have an interview with Mahzarin Banaji coming up -- WHAT would you really love me to ask her? Prasad, what would you love me to ask her?
Posted by: Elatia Harris | Mar 3, 2013 10:14:29 PM
Here are some questions I'd love to hear Banaji tackle:
1. Adress Levitin's criticisms (http://online.wsj.com/article/SB10001424127887323829504578272241035441214.html?mod=wsj_share_tweet) . Especially:
a. The IAT does a horrible job in predicting real world reactions
b. Reaaction to Banaji's own meta analysis that showed the IAT in a very poor light
2. What does Banaji think of Bargh's work
I am left amazed that this kind of work passes for science. And that too at a major university.
P.S. A funny anecdote: Malcolm Gladwell is racists according to the IAT.
P.P.S. I don't want to pick on Elatia's choice of words, but when one reads something like: "But this is a good example of how hiring policies (practice) can be ahead of testing (theory)." , it is natural to wonder what exactly Elatia's attitude to this work is.
Posted by: Sundar | Mar 4, 2013 6:03:53 AM
Thanks, Sundar -- I think. But. you seem to be on a misconstruing streak. My attitude towards this work is very plain: it is fascinating. I don't put my time into things that are not, to me, extremely worthwhile.
The classical music world's retreaded auditioning protocols that I mention are absolutely the result of becoming conscious of a serious bias and countering it in practice, leading to a result that makes a big difference -- a huge increase in women orchestra musicians. This was done by blind auditions, not by social engineering. But the powers that be had to accept that their bias was a indeed bias, not a gender-blind assessment of the auditioning musicians' quality that just happened to overwhelmingly favor men. It began with becoming conscious of the bias, and deciding NOT to to seek some other explanation for the bias, but acting to remove its impact from their hiring practices.
The IAT is a powerful tool for discovering and confronting bias in yourself and your institution. But you have to be open to its message if you are to resolve to act in ways that stop the impact of your bias in its tracks. Insight and action are not always found together. You can for instance respond with a few minority hires. But this is not what the people in my example did -- they found a way for the bias to be made powerless, and hired based on blind auditions. But it all began with a look in the mirror, don't you think?
Posted by: Elatia Harris | Mar 4, 2013 8:10:01 AM
Elatia, you write: "The IAT is a powerful tool for discovering and confronting bias in yourself and your institution. "
Why are you so certain that this is the case? I find Levitin's criticisms of the IAT very convincing. Don't you?
Posted by: Sundar | Mar 4, 2013 9:48:28 AM
I find those criticisms an interesting aspect of the dialogue, and not more. But why ask me? You can take the test for free, and ponder whether you feel it's accurate in your own case. If you discover a slight bias, then you might want -- genuinely want -- to take corrective steps of your own choosing. It won't wave a wand, but it will help you keep mindful of the difference between intuitions about individuals you may meet and biases towards or against the groups they represent. This is why I would call the IAT a powerful tool, and not a powerful solution.
Posted by: Elatia Harris | Mar 4, 2013 2:31:15 PM
Elatia, I think the points in the WSJ review are good ones (Philip Tetlock has made very cogent critiques of IAT as well) so I hope some of those points will come up during the interview.
For my own part, most of the stuff I want to ask Banaji revolves around the extent to which she simply wants to affect our _beliefs_ rather than our _attitudes_ by blinding us to statistics. So here's the issue I'd to see her pressed on. In the bloggingheads interview, Banaji is asked directly why some of her 'implicit association' results don't just amount to people being aware of social facts about the world. She replies that in fact people have really bad intuitive grasp of base rates. That's true, but:
A) the directly relevant question Banaji should have asked for a contrastive association test is "do people have a good intuitive grasp of likelihood ratios" not "do they have a good grasp of base rates." To do a classic test case, if I feel my pulse racing when I see a stranger in an alley way, base rates affect whether this fear is sensible. But if I am _more_ scared (which is what IAT is all about) when the stranger looks poor, the relevant question is whether I have a decent grasp on the ratios of risk posed by rich/poor people.
I think this question doesn't have the answer Banaji needs - I think the research shows that people are lousy at estimating absolute risk, but are actually pretty decent at assessing relative risk. You can play the same game with race too, though Banaji has an extra handle there, viz that if you're prejudiced against a group you're especially bad in estimating its base rates.
B) Her response is a dodge. If you say people don't know base rates, the *Bayesian* solution is to say "let's become better acquainted with statistics and facts." The Banaji response is to say "let's change our mindsets to think less of stats and more about being the change we wish to see." Those are NOT the same prescription.
Basically, I keep circling around this: Banaji's call seems to be "don't know/ignore facts that might inappropriately affect your attitudes vis-a-vis your political goals." This seems like at least a potentially iffy approach.
Anyway, looking forward to your interview.
Posted by: prasad | Mar 5, 2013 1:54:33 PM
I have not listened to the Banaji interview, but Prasad's description confirms my assessment of the whole project. Especially Prasad's description:
"The Banaji response is to say "let's change our mindsets to think less of stats and more about being the change we wish to see."
Elatia, You seem convinced that the IAT measures something called "bias". I on the other hand think it measures nothing at all.
It has no predictive power. It does not agree with other measures (as Banaji's own meta analysis shows).
It is basically a handy pseudo-scientific tool to advance political agendas aka "being the change WE wish to see"
Posted by: Sundar | Mar 5, 2013 4:24:56 PM
Well, guys, science exists within culture, not the other way around. This test was developed in the era we live in, and it reflects that context.
There is such a thing as the subjective experience of taking the test. And that's a little different from the subjective experience of gaming the test, or of testing the test.
I would advise everyone wondering about the test to just take it. See what it tells you. Otherwise, you are judging a wine by reading about it, looking at it, checking whether the vintner is kosher, and chatting about it with others who have also not tasted the wine in question. You may not be wasting your time, but you are taking an elliptical approach to subject.
Posted by: Elatia Harris | Mar 5, 2013 6:05:15 PM
I agree completely with Elatia re taking the test. It's well worth taking, and is a pretty informative experience. Also, it's fairly plausible a priori to think rapid reactions might probe instinctive responses - it's similar to making people answer questions rapidly to avoid deception or spin. Also I don't need that much convincing that it's getting at something important - my priors are not much different re _existence_ of prejudices etc.
Of course I do also think saying that is compatible with saying as well that Banaji's research needs careful scrutiny. More so frankly because her work basically tells her colleagues and audience exactly what they want to hear. So the audience needs to guard against wishful thinking too.
Posted by: prasad | Mar 5, 2013 10:44:23 PM
Post a comment