Invisible Bias

By Chris Berdik

Chris Berdik is a freelance journalist in Boston. This article appeared in the Boston Globe on December 19, 2004. Reprinted with permission.

June 2005

A group of psychologists claim a test can measure prejudices we harbor without even knowing it. Their
critics say they are politicizing psychology.

Inside the wood-paneled confines of the Harvard Club, about 200 Bostonians gathered recently to
tap into their subconscious. Literally. Audience members were told to move as quickly as possible
through a series of faces and words projected on a screen, tapping their left knees for a young
face or a “good” word (joy, sunshine, love), and their right knees for an old face or a
“bad” word (bomb, agony, vomit). It took about 15 seconds for most to finish. But when asked to switch,
to pair young faces with “bad” words and old faces with “good” words, the rhythm faltered and the tapping slowed. Audience members shook their heads and giggled. Some threw up their hands.

To the Harvard psychologist Mahzarin Banaji, who presided over the event, the demonstration
showed that most of the audience—like most of the people who have been subjects in this type of
experiment—have a harder time associating old people (or nonwhite people, or homosexuals) with “good”
when given no time to think. These are all examples of what Banaji calls implicit prejudice, and their
importance extends way beyond an intellectual parlor game. Implicit prejudice, she argues, can affect our
decisions and behaviors without our even knowing it, undermining our conscious ideas and best intentions
about equality and justice.

Such implicit prejudices are “ordinary,” says Banaji. “Ordinary people show them. They stem
from ordinary cognitive processes.”

About a decade ago, Banaji and Anthony Greenwald, a psychologist at the University of
Washington, developed a test for uncovering these subconscious preferences—the Implicit Association
Test (IAT). Normally, instead of tapping knees, an IAT subject uses a computer keyboard to group “good”
and “bad” words with images as split-second differences in response times are measured and tabulated.

Today, some 8,000 people a week take an IAT on the website of Project Implicit (https://implicit.harvard.
edu/implicit/), founded in 1998 by Banaji, Greenwald, and Brian Nosek, a University of Virginia psychologist. The site has dozens of tests measuring implicit biases on everything from politics to race to gender roles. Some results so far: 75 percent of white respondents implicitly favor white over black, more than 70 percent of all respondents favor straight people over gay people, and about 80 percent favor young over old.

To Banaji and a growing number of researchers, the IAT has potential uses far beyond the lab. This
year, Banaji is heading a group of psychologists and legal scholars at the Radcliffe Institute to develop new
approaches to anti-discrimination law based on the idea of implicit prejudice. The IAT has been proposed
for use in corporate ethics classes, police and other professional training, and in consumer research.

But not everybody trusts the IAT. Social psychologists are divided on just what the IAT measures, arguing
that different response times may just reflect an awareness of cultural stereotypes and social inequality.
In February, the journal Psychological Inquiry will devote an entire issue to the debate surrounding the
test. And beyond the technicalities, a bigger question looms: If prejudice really is rooted deeply in our
subconscious minds, how can we get rid of it?

The foundation for a social scientific study of prejudice was laid 50 years ago by a Harvard
psychologist named Gordon W. Allport in “The Nature of Prejudice.” Prejudice, Allport wrote, grew
from the instinctive way people simplify their world by categorizing everything—including other people.

According to Allport, we have various automatic expectations based on probabilities. We assume, for
instance, that a man in a three-piece suit has money and employment or that the person sitting beside us
in church shares our basic beliefs. Allport noted that while such expectations aren’t always correct,
they’re useful and generally harmless. For Allport, prejudice—the dangerous phenomenon that could
lead to everything from racial slurs to lynchings— began when those expectations were accompanied
by conscious antipathy toward a particular group and were inflexible in the face of contradictory evidence.

Allport’s treatise remained a foundation for psychological research into prejudice for decades.
Indeed, Banaji and her colleagues begin with the premise that prejudice has its roots in the normal
human tendency to categorize. But they veer sharply from a fundamental tenet of Allport’s theory. In their
view, you don’t need to have antipathy toward any particular group to harbor implicit prejudices that
could lead to discriminatory behavior. Instead, according to IAT researchers, implicit prejudices
build over time as stereotyped images seep into our brain—news images of the African-American suspect
or the Arab terrorist, commercials where wives clean the house, the not-so-bright sitcom character with a
Southern drawl.

Says Banaji, “Seeing is believing, at least at some level.”

A big reason for the persistence of these prejudices, she emphasizes, is denial. People with
strong egalitarian values know there are prejudiced people out there who act in prejudiced ways, but
they don’t allow that they might be one of them. Banaji argues that this denial is rooted in the desire
to believe that our judgments and actions are all within our conscious control.

IAT co-creator Greenwald agrees: “There are many, many well-meaning people who attend diversity
trainings and say, ‘I’m happy to go along with this, but it’s not my problem.’ But with the IAT, people
discover, ‘Well, there’s something going on in my head, too.”

That’s why Banaji and her colleagues at the Radcliffe Institute think it’s problematic that much
anti-discrimination law requires plaintiffs to prove an employer or other individual intended to discriminate.
They hope to spread the idea of implicit, unintentional prejudice throughout the criminal justice system.
And they hope to develop legal arguments, bolstered by theories of implicit prejudice, that could prove in
court that an employer’s hiring and promotion policies discriminate against women or minorities, for example, even without any conscious intent.

While Banaji says many subjects react negatively to being told they exhibit implicit prejudices, those at
the Harvard Club who cared to comment after the presentation seemed convinced. “I think everybody
has biases. It’s part of being human,” said va Das, 62, a civil rights lawyer. “I think the only real question is
what to do about them.” Bob Frankel, a 59-year-old research engineer at MIT added, “I think one of the
values of people taking tests like [the IAT] is so they realize, ‘OK, maybe I’m not quite who I thought I
was.” But not everyone thinks Banaji and her colleagues have necessarily discovered a hidden
reservoir of prejudice. The dissenters, a number of whom have articles in the upcoming issue of
Psychological Inquiry, argue that a speedier association of white with good and black with bad may simply
reflect a subject’s awareness of societal inequalities, such as the disproportionate number of blacks in
prison, rather than a subconscious bias.

The principal critique of the implicit prejudice theory, written by Hal Arkes of Ohio State and Philip
Tetlock of Berkeley, carries the subtitle “Would Jesse Jackson ‘Fail’ the Implicit Association Test?” In one
section, they speculate whether Jesse Jackson and Jesse Helms would score similarly on the IAT.

“Although the two figures disagree profoundly on certain political issues,” the authors note. “They
agree that the ‘African-American family’ is in trouble, that African-American crime rates are far too high,
and that African-American test scores are too low. Should we theoretically expect indices of ‘negative
affectivity’ [such as the IAT] to differentiate people who share a considerable knowledge base but who
differ only in their causal attributions for betweengroup inequality?”

Instead, Arkes and Tetlock argue that to conclude a person is prejudiced, one should stick
with the Allport standard, which says that prejudice requires some level of hostility toward a particular
group. What’s more, they say, Banaji and other IAT promoters are “politicizing” psychology. “We suspect
that, when the history of social psychology is written at the end of the 21st century,” they write, “implicit
prejudice research will be a prime exhibit of how society became so obsessed with avoiding stereotypes
that it skewered citizens as racists for displaying even trace awareness of politically painful realities.”

But Banaji dismisses the argument that the test simply reflects “awareness” of stereotypes and
inequalities. She brings up a recent “meta-analysis” of more than 60 studies that show the IAT to be a
better predictor of behavior than explicit measures of attitude in sensitive areas such as racial interaction.
Among white subjects, for example, a strong subconscious bias for whites over blacks among
white subjects was correlated with behaviors such as lack of eye contact with a black test administrator.
Other IAT lab experiments found implicit prejudices correlated to more negative ratings of a black author’s
essay and a greater willingness to make hypothetical cuts in the budgets of minority student groups.

“If it’s just an activation in my head, if it’s not my attitude, then it shouldn’t affect my behavior,” says
Banaji. “We would all agree that this is something that comes from the culture. But I would say it becomes us.”

Banaji recently bought some postcards featuring prominent people of color: Jackie Robinson, Zora
Neale Hurston, Ghandi. She scanned them into her office computer, and they now cycle through as
screen savers. It’s part of the ongoing effort of this Indian-born psychologist to rid herself of her own
pro-white IAT bias. “[My race bias] troubles me perhaps more than any other one,” she told the
audience at the Harvard Club. “I try to beat that test all the time.” Banaji also admits to other implicit
biases, such as associating men more strongly with careers and women with the home.

Indeed, Banaji and fellow IAT researchers are investigating ways of mitigating the biases their tests
uncover. Some methods are more passive, such as altering the environment where we live and work to
increase exposure to images and situations that contradict prevalent stereotypes. Experiments reveal,
for instance, that having an African-American administer the test to subjects lessened their
pro-white bias, as did having a subject view images of admired African Americans just before taking the
test. (In fact, black IAT subjects are split almost evenly between favoring black and favoring white.)

But the researchers believe the remedy can’t just be passive. “Just as we need to do work with the physical body, I think mental muscles need the same kind of treatment,” says Banaji.

At least one member of the Harvard Club audience seems to have taken this advice to heart. After
Banaji’s lecture Sarah Smith considered “the necessity of meditating on people who are not like me…to
stretch my sense of who I’m in the same human boat with, as it were.” As an example, the 56-year-old writer from Brookline quipped, “I now try thinking kind thoughts about Republicans on a regular basis.”

Writing on the cusp of the civil rights movement, Allport noted, “It required years of labor and billions
of dollars to gain the secret of the atom. It will take a still greater investment to gain the secrets of man’s
irrational nature. It is easier, someone has said, to smash an atom than a prejudice.”

In the decades since, much overt or sanctioned discrimination has been eliminated from American
society. Congress passed the Civil Rights Act of 1964, the Voting Rights Act of 1965, and the
Americans with Disabilities Act of 1990. But the scholars who debate the IAT give this progress
different spins. In Psychological Inquiry, Arkes and Tetlock write that it’s disconcerting that “cognitive
research programs now attempt to gauge prejudice not by what people do, or by what people say, but
rather by milliseconds of response facilitation or inhibition in implicit-association paradigms.” In
another paper they ask, “How far down the continuum should we venture in pursuit of ever-sneakier forms
of racism?”

On the other side, IAT researchers point to the concern many feel about the persistence of social
inequalities. Minority populations continue to have higher rates of poverty and unemployment than do
white populations. A racial “achievement gap” persists in education. Women still earn only about 80 cents
for every dollar earned by men.

“It’s difficult to be optimistic given that all that well-meaning activity of the second half of the 20th
century hasn’t been able to rid us of these disparities,” says Greenwald. He adds, “I think if we can use the
IAT to spur a new look at human attitudes and stereotypes, then we can begin to develop a new
model of the person and educate people about it. And I think this is the basis of some optimism.”

Banaji insists that these efforts should be more about awareness than about guilt. “My job is not to
construct ethical theories or define the ultimate good,” she says. “We’re interested in revealing to
people that their own moral and ethical standards are being compromised by the stuff in their heads.”

Back to June 2005 Contents

Back to STATUS Table of Contents