The Science behind Implicit Bias
Are you biased against women leaders? Most people reading this article would quickly and resolutely answer no. But when AAUW posed that question to members and supporters in February, researchers found a different answer. It turns out, most people still associate leadership with men more strongly than they do with women.
Of course, no one expressed those opinions outright. Rather, they were the results of an Implicit Association Test (IAT) created by AAUW and Harvard’s Project Implicit team to measure the subconscious biases that help explain why, despite all the strides women have made, they’re still woefully underrepresented in the highest leadership positions.
Developed in 1998 by Anthony Greenwald of the University of Washington, Mahzarin Banaji of Harvard University, and Brian Nosek of the University of Virginia, the IAT asks test takers to sort words into categories and measures how long it takes to make a correct categorization. A typical IAT instructs participants to sort words or images as quickly and accurately as possible into categories that range from the innocuous flower/insect, good/bad to the more loaded black person/white person, good/bad or, on the AAUW test, men/women, leader/follower. The researchers who created the test argue that a faster sorting speed indicates an easier mental association between categories. The IAT compares your individual responses only with your own results, measuring the relative differences between your implicit associations.
“Basically, to the extent that people are faster at doing categorizations when white is paired with good and black is paired with bad is an indirect way of assessing their association between white and good and black and bad, their implicit attitude,” says Kate Ratliff, an assistant professor at the University of Florida and executive director of Project Implicit, a nonprofit collaboration among researchers working on implicit social cognition.
Why the Unconscious Mind Matters
“These sorts of tests allow us to get at the things that sometimes people literally don’t even know that they know,” says AAUW Senior Researcher Kevin Miller, who helped develop AAUW’s gender and leadership IAT. Although people may be quick to disavow old-fashioned or prejudiced notions, their subconscious mind is still subject to the influence of their environment. In the United States, that environment often still tells them that women are bad at math, and only men are presidents. And it’s those implicit associations, not conscious values, that dictate how most people treat anyone who isn’t a tall, thin, conventionally handsome, white, cisgender male.
“There’s been a lot of research done that correlates people’s scores on a test like the IAT with their actual behavior,” says Ratliff. “And we see pretty strong correlations. Or at least we see correlations that are stronger than if we try to correlate their self-reported attitudes with their behavior.”
In the book Blindspot: Hidden Biases of Good People, IAT creators Banaji and Greenwald list troubling behaviors accurately predicted by race IAT scores. These include preferring white job candidates to identically qualified black candidates in simulated hiring scenarios and doctors prescribing the “optimal treatment” for a given condition more often to white patients, despite their presenting the same symptoms as black patients.
Imperfect Humans, Imperfect Science
Like what you’re reading?
This article originally appeared in the Spring 2016 of AAUW Outlook magazine.For more stories like this, subscribe to Outlook today.
Banaji and Greenwald call these errors in judgment “mindbugs,” short circuits in the otherwise extraordinarily efficient and effective cognitive processes of the human brain. But so far, social scientists know only that mindbugs happen, not necessarily how or why.
The IAT can’t explain the bias; it simply detects it. But it’s far from a perfect measurement, in part because the human brain is notoriously susceptible to suggestion. One study described in Blindspot asked subjects questions about admirable black men such as former Secretary of State Colin Powell and white criminals such as “Unabomber” Ted Kaczynski before they took a race IAT. The subjects showed a weaker association between white and good than did the control group.
“In an ideal world, you’d use a tool that could just take a snapshot of someone’s brain and tell us their exact level of bias, but each tool only gives you a limited slice,” says Miller. And that slice is “influenced by a wide variety of things: the setting, the other things you’ve done that day, if you’re thinking about what you’re going to do next, and so on.”
The test has other limitations as well. It can only evaluate two pairs of categories at one time so that, for example, women can be compared with men, but other differences can’t be taken into account at the same time. The test also still requires a computer keyboard to complete, though Ratliff says that a mobile version is in the works.
And, perhaps most crucially, the test is created by humans, which could explain why an older gender and leadership IAT, created by a different organization, paired leader words such as “dynamic” and “assertive” with so-called follower words like “compassionate” and “understanding,” two words that may indeed describe followers but that also carry gendered connotations. (Never mind the fact that “compassionate” is a word that describes some of the most successful leaders.)
“Some of the words on the list didn’t really seem like fundamental traits of someone who follows. In many ways they seem like fundamental traits, in our culture, of femininity. And so we thought the lists were problematic,” says Miller.
That meant that coming up with new, more neutral words for the categories of male/female and leader/follower. So Miller and the other researchers compiled a list of synonyms for leader and follower instead of traits that describe leaders and followers.
The AAUW researchers also had to choose male and female names to sort on the test, so they zeroed in on the 25 most common men’s and women’s names in the United States, with a focus on names with an average age of 35–55 years old in order to avoid names that might be too linked to a younger or older generation. That list had its own problems: “Donald was one of the most common names,” says Miller, so it had to be dropped because its association with a presidential candidate might skew the results. “Hillary wasn’t one, but if it had been we would have dropped it,” he adds.
Now That You Know, What Now?
So how do people who consciously reject their unconscious bias counteract it? The answers are still unclear.
“We think of awareness of bias as being the first step,” says Ratliff. “It’s a completely necessary step on the path to preventing bias from influencing your behavior, but on its own it certainly isn’t enough.”
Ratliff and Miller both point to blind musical auditions, which dramatically increased the numbers of women and people of color in symphonies. The spirit of this practice can be invoked by removing identifying details from résumés or other applications and by setting concrete, objective standards where possible.
But the real first step may just be admitting you have a problem. As part of her work, Ratliff studies respondents’ reactions to their IAT scores. “You might not be surprised that people are more defensive when they get a score that they don’t want than they are when they get a score that they do want,” she says.
So what would she tell a group of feminists who just found out that they express a slight bias toward male leadership? “The biggest thing to keep in mind,” Ratliff says, is that the tests “don’t reflect what we want, they don’t reflect what we consciously believe, so … they’re going to contradict our values and our beliefs. Men are in leadership positions more often, men are stereotyped as being leaders more than women are, and I think to a larger extent our scores on these tests reflect that.”
Elizabeth Bolton was an editor at AAUW for eight years before becoming a brand strategist and project manager for a leading design firm. Her IAT results show a slight association between women and leadership.