Starbucks WIll Give Employees Unconscious Bias Training. That May Not Help.

After the manager called the police and the two black men waiting to talk business with a friend in a Starbucks were arrested Friday, the company kept apologizing.

First on Twitter Saturday:

Then in a statement from CEO Kevin Johnson lamenting “a disheartening situation … that led to a reprehensible outcome.”

Then Johnson a second time on Monday’s Good Morning America:

Then the chief operating officer, Rosalind Brewer, who is black, said Monday on NPR that the company planned to respond by training its employees about unconscious bias:

But what is “unconscious bias training”? Or rather, what is “unconscious bias”?

“Good question,” Calvin Lai, an assistant professor in the department of psychological and brain sciences at Washington University. Lai has studied implicit bias and published papers in 2014 in the Journal of Experimental Psychology and in 2016 in the same journal that explored a central tenet of the Starbucks case: Can those of us who have inherent racial biases control those urges?

Implicit bias refers to stereotypes we resort to—unconsciously—that affect how we interpret a situation and the characters within it. Lai said academics and the greater public have diverging thoughts on what counts as implicit bias and what doesn’t.

“A lot of people talk about implicit bias to refer specifically to things like race bias and gender bias, whereas a lot of people use unconscious bias to refer to biases ... in terms of how we make our decisions in everyday life.” But Lai said in the end, it’s “jargon wars” and that biases often get clumped together into “implicit bias” when companies like Starbucks try to address sticky situations like the Philadelphia incident.

Beneath the umbrella term of implicit bias, there are several branches—not all agreed upon by social psychologists. “If you ask 100 different social psychologists [the difference between unconscious bias and implicit bias], they’ll give you slightly different definitions, each and every one of them,” Lai said.

And while the word “implicit” might invoke some sense of the bias being unconscious, that’s not true. Lai said that in fact, there are four biases that fall under implicit bias: “Are people aware (so consciousness or unconsciousness); how controllable or uncontrollable it [the bias] is; how quick it is to arise... like snap judgments compared to slower biases; and how efficient it is, [or] how much mental effort it is to do or think something.

“Those four things can be unrelated to each other,” Lai said. “When we’re talking about unconscious bias, we’re talking about one very specific type of automaticity, or implicitness.”

Implicit—or unconscious bias training—all seems magical: Gently show befuddled humans that they’ve got some sort of bias—whether it be racial or gendered or regarding sexual orientation—and make them aware of those thoughts so that they are empowered to not act without conscious intent in a bigoted way.

You know the saying: If it’s too good to be true, it probably is. And when it comes to implicit bias training, the science just doesn’t hold up: There isn't much scientific proof that it actually works.

That hasn’t stopped implicit bias training from being incorporated into the human resources manuals of major American corporations, universities, Silicon Valley strongholds like Google and Facebook, and hip startups that have adopted the practice and boasted of using it as proof the companies were inclusive and diverse. Most important, these bias training workshops give the impression that companies are addressing discrimination head-on, giving agency to employees to address their own personal prejudices within the privacy of their minds.

That, in fact, is what makes figuring out the Philadelphia Starbucks situation that much more complicated for social psychologists. Lai said that as a scientist, it’s hard for him to analyze a video and know if either unconscious or implicit bias is at play because we don’t know anything about what either party is thinking—or rather, not thinking.

“All we know about the behavior of the men is what is exhibited on the tape,” he said. “Considering that social scientists collect data on hundreds of people and have trouble coming to a conclusion from even that, we have no idea what happened in this case.”

What complicates matters further is that defining racism is something social scientists haven’t quite universally agreed upon, Lai said. What's tricker is that the video doesn’t show visual or verbal signs of racism. While it’s nearly universally agreed that calling the two men “trespassers” was extreme, it's hard to figure out what motivated the phone call to police in the first place.

So what does “unconscious bias training” even mean? Lai said several groups had tests organized that allowed people to assess whether they had implicit biases ingrained. The most famous test is the Implicit Association Test, or IAT, a Harvard nonprofit that’s been administered to tens of millions of people. It operates by having users organize words quickly—with as little thinking as possible—into classifications. “The IAT measures the strength of associations between concepts (e.g., black people, gay people) and evaluations (e.g., good, bad) or stereotypes (e.g., athletic, clumsy),” the site describes the test. “The main idea is that making a response is easier when closely related items share the same response key.”

But that’s vague wording, and testing has been all over the place. “I can name all the rigorous experiments [on implicit bias training] on one hand,” Lai said. “That’s not saying that they don’t work and that other diversity-type training is better. It’s just that we don’t know and that there isn’t enough research.”

That’s because diversity training hasn’t been standardized—and it’s possible it might not be able to be standardized. Measuring prejudice, particularly that which might be deeply ingrained within our psyche and outside our knowledge of its existence, is profoundly difficult to recognize and measure. A word association test like that of the IAT might give us a vague understanding of a person’s worldview, but it might also hide deeper issues of bias within itself. As an extensive analysis in Quartz points out, the test-retest reliability score is crappy. In plain English: Retaking the test often resulted in different results, which means the test was inconclusive and/or faulty in measuring what it set out to measure.

And there's some murkiness in even connecting the fact that a person has implicit bias and actually actually acts on it. Over the past few years, a series of studies have found weak correlations between implicit bias and behavior, finding time and time again that implicit bias training might not live up to its promises.

Lai’s suggestion that bias training is one big shrug-worthy exercise hinges on the fact that his previous papers on the subject—which considered two interventions designed to try to help people with racial biases identify them before acting upon them—found that effects lasted up to 48 hours. That’s right: Bias training seemed to affect subjects for a couple days before they reverted to their more automatic way of connecting stereotypes with race. Aside from recognizing these biases, people were not likely to change their behavior.

It's important to note, Lai said, that these experiments were focused on changing implicit bias directly rather than the type of implicit bias training an organization might run. But still, they bring up questions about the process of identifying—and addressing—implicit bias.

And it doesn’t help that most organizations don’t personally involve themselves in measuring implicit bias, instead incorporating it into an HR series for employees to hurriedly click through and complete. “I was taking one of those modules for HR the other day,” Lai illustrated as an example, “and they kind of throw the information at you with no training or context. No offense to our HR, but it’s not that engaging. Sometimes, they [other companies] have a person educate people about bias. You think about it, but that’s it.

“It’s a little like telling people they have high blood pressure without telling them what to do about it.”

So could a program that Starbucks is putting forth, the vaguely lofty unconscious bias training, actually work? Maybe. Is there proof it does? Lai flatly said no.

“There isn’t good evaluation, period,” he said. “For one reason or another, there hasn’t been any academic research, and part of that is it’s difficult to get access to a company and to collect indicators that might make them liable.”

Legal issues aside, Lai said the stickier, murkier part of implicit bias training is recent evidence that not only might it not work, it might hurt. In 2016, Alexandra Kalev and Frank Dobbins published research in the Harvard Business Review analyzing such diversity training initiatives between 1985 and 2000 and their after-effects. What they found was counterintuitive: Mandated diversity training often reduced hiring and promotion of minority individuals.

“It’s counterproductive, they found,” Lai said. “You bring these managers in and they’re feeling pissy and reactive. So they’re less motivated to hire minorities in the future.”

It becomes a vicious circle of groupthink, one that ultimately creates groups of people who look and think and believe similarly because the people they are with are just like them.

That’s not to say Lai thinks implicit bias training is an absolute waste of time. “There’s two ways of thinking about this,” he said. “Corporations are taking the easy way out [by implementing implicit bias training], or, well, if there’s an easy thing to do [like implicit bias training], then why not do it?” And an advantage that can’t be denied, Lai said, is that people at least pick up on what their biases are.

As of publication time, a request for comment from Starbucks about what its unconscious bias training entailed went unreturned.