Implicit Bias

Implicit association, or implicit bias, is best explained through analogy. Suppose I handed you a deck of standard playing cards, and asked you to quickly sort the cards by suit, with hearts and diamonds in one pile and clubs and spades in another. Readers who easily distinguish red from black would find this easy, and make few mistakes, because our minds find it easy to sort by color. It’s a more difficult task to quickly sort hearts and clubs (red and black) into one pile, and diamonds and spades (also red and black) into another: the task takes longer, you find yourself hesitating over some cards in a way you never did when sorting by color, and your performance is less accurate. The problem, of course, is that this second task requires us to defy the simple organizing principle of sorting by color that made the other task so simple.

Researchers, including Mahzarin Banaji and Anthony Greenwald, authors of Blindspot: Hidden Biases of Good People, have presented test subjects with a similar task with graver consequences. Instead of sorting playing cards by suit, subjects sort other categories: pictures of Caucasian faces, pictures of African American faces, words with positive associations (happy, joy) and words with negative associations (angry, sad). (You can take the test yourself, as well as many other implicit association tests, by visiting implicit.harvard.edu.) What these psychologists have found is striking: around 70% of American test subjects (including roughly similar proportions of both Caucasians and African Americans) associate Caucasian faces with positive words, and African American faces with negative words, much more readily than they associate African American faces with positive words, and Caucasian faces with positive words. Precisely how this comes about is unclear, with some suggesting that media coverage or vicarious experiences contribute to the phenomenon, but even without understanding the mechanism, the results appear to be robust. The researchers’ inference is that, regardless of race, most participants implicitly have positive associations with Caucasians, and negative associations with African Americans; this phenomenon is called implicit association or implicit bias. Greenwald published his first data on the subject in 1998, and the experiment has since been repeated elsewhere. For example, it appears that, under experimental conditions calling for rushed judgments, such as a video game, many subjects will more readily perceive a weapon in the hand of an African American (even mistakenly so), and a wallet or cell phone in the hand of a Caucasian, than the reverse, suggesting an implicit perceived association between race and violence. In a different experiment, test subjects were ostensibly charged with a routine counting task on a computer. Unbeknownst to them, the computer program was designed partway through the process to momentarily (and subliminally) display a human face and then crash and restart, and their reactions were recorded. No test subjects were, we are given to understand, consciously aware that a picture flashed across the screen. However, test subjects who were shown an African American face tended to exhibit more anger and frustration with the computer crash than those who were subliminally shown a Caucasian face.

It is crucial to recognize that the implicit association test does not reflect test participants’ conscious attitudes. Many, even most, test participants to whom the implicit association test attributes negative associations with African Americans (Banaji and Greenwald among them) report that they feel positive attitudes toward African Americans, and are upset by what the test reveals. For many, the results of the implicit association test suggest implicit bias against groups that include their parents, spouses, children, or close friends. We can possibly infer from this that investing time in personal relationships can overcome implicit bias in individual cases, but although mildly comforting, this, like the “some of my best friends” argument, misses the larger point. Implicit bias suggests that many people experience a default, unconscious sense of aversion toward members of some racial groups but not others. Because of implicit bias, members of less preferred groups discover to their immense frustration that they have to overcome a default attitude of aversion to be eligible for the same benefits and opportunities that others, born into a comparative state of favor, simply take for granted. Even people who are strongly committed to racial equality would do well to examine their implicit associations.

On the other hand, I would argue that in situations that do not require hasty judgments – including nearly all employment decisions – managers’ basic moral judgment and familiarity with anti-discrimination policies will generally prevail over whatever instinctive, unconscious reaction we might experience as a result of implicit associations. (Greenwald himself, testifying in 2012 as an expert witness for the plaintiffs in the class action employment lawsuit titled Pippen v. State of Iowa, refused to conclude that implicit bias had been the cause of any difference in hiring of white and black employees by the Iowa state government, and said he could not rule out race-neutral reasons for those decisions. The court further held that the plaintiffs’ allegations of implicit bias were insufficient to establish a causal connection between supervisors’ discretionary and subjective hiring decisions and the statistical hiring disparities that plaintiffs had identified.) Unconscious attitudes are not the extent of our character; so for all but the most impulsive and unreflective managers, implicit bias may have little bearing in established employment relationships.

Unfortunately, I’m not sure we can say the same thing about encounters between strangers. The versions of the implicit association test calling for subjects to distinguish weapons from innocuous items would seem to have immediate application in law enforcement, and Banaji and Greenwald recount an alarming anecdote about how implicit bias appeared to affect the course of medical treatment in a prestigious hospital. They argue that implicit bias has some of its most significant effects not in circumstances where laws or social norms govern our behavior, but in unregulated encounters, such as the decision whether to go out of our way to assist someone else: giving directions, offering mentoring, offering medical assistance, and other chance meetings between strangers.

Just as unfortunately, it doesn’t appear that any overt method has been developed that can remediate the tendencies identified by the implicit association test; Banaji and Greenwald confess to each retaking the implicit association test many times, trying (and failing) to improve their scores, and Greenwald testified in the same Iowa lawsuit that training would not necessarily prevent implicit bias from affecting conduct. This is not to suggest that scientists have not bothered to explore remedial strategies. Nilanjana Dasgupta and others have devised experiments that administered the implicit attitude test after exposing test subjects to images of highly respected African Americans and infamous white Americans; Banaji (though without expecting therapeutic effects) runs a screen saver that cycles through positive, stereotype-disrupting images to remind her of humanity’s diversity. Although reminding test subjects of positive role models in these ways may reduce implicit bias temporarily, the effects dissipate; researchers have not yet developed a technique that can suppress implicit attitudes over the long term.

However, an indirect approach, starting with self-awareness, can also be helpful in coming to terms with subtle habits of thought or behavior. Readers interested in developing greater self-awareness may want to take the implicit association test themselves, and reflect on any discrepancies the test reveals between their social views and their cognitive wiring. Beyond that, they may want to notice a few more of life’s details: Am I smiling? Making eye contact? Blinking often or seldom? How far away am I choosing to sit or stand from the other person? Where the occasion presents itself, am I choosing physical contact (a handshake, handing the object to the person directly) or keeping my distance? Am I choosing to be engaged? Am I choosing to imagine that person’s point of view? Do I find myself becoming nervous, agitated, or self-conscious? Am I exploring a social opportunity, or avoiding one?

In the privacy of our own thoughts, there’s no need to become defensive about what we might notice from such experiments. However, if we’re in the habit of attending to such things, we may discover that social situations present us with choices we hadn’t previously been aware of.