In response to online and in-person protests at Starbucks after a manager at a Philadelphia cafe called the police on two black men last week, the coffee giant announced Tuesday that they will be closing over 8,000 locations in the United States on May 29 for company-wide racial bias training.

“I’ve spent the last few days in Philadelphia with my leadership team listening to the community, learning what we did wrong and the steps we need to take to fix it,” said Starbucks CEO Kevin Johnson in a statement. “While this is not limited to Starbucks, we’re committed to being a part of the solution. Closing our stores for racial bias training is just one step in a journey that requires dedication from every level of our company and partnerships in our local communities.”

While Starbucks has yet to release details about the racial bias training, the statement said that the training program will be "designed to address implicit bias, promote conscious inclusion, prevent discrimination and ensure everyone inside a Starbucks store feels safe and welcome."

Sounds good, right? But there's a problem.

As Jesse Singal wrote in New York magazine last year, "Perhaps no new concept from the world of academic psychology has taken hold of the public imagination more quickly and profoundly in the 21st century than implicit bias—that is, forms of bias which operate beyond the conscious awareness of individuals." The idea, broadly, is that all people (white, black, brown, etc.) hold subconscious bias against certain demographics, and this bias can be reflected in Implicit Association Tests, which were developed by Harvard University psychology chair Mahzarin Banaji and University of Washington researcher Anthony Greenwald.

If you haven't taken the test, it works like this: You are shown a series of images and asked to pair them with a series of words that have either strong positive or negative correlations. For instance, I just took the test for sexuality bias, and positive words (Laughing, Happy, Friend, Friendship, Glorious, Attractive, Adore, Cheer) were contrasted with negative words (Grief, Sadness, Awful, Abuse, Selfish, Bothersome, Rotten) as well as images of either two blocky female figures (of the bathroom sign variety), two male figures, or one male and one female together. My task was to pair the images with the words, and according to the test, based on how quickly I sorted them, the test determined that I show "a moderate automatic preference for Gay people over Straight people," which, frankly, I didn't need a test to point out.

This test is hugely popular, in no small part thanks to Malcolm Gladwell, who wrote about it in his best-selling book Blink: The Power of Thinking Without Thinking. Gladwell called the IAT "a powerful predictor of how we act in certain kinds of spontaneous situations," and, in the years since its inception, the IAT has been taken many millions of times and has been incorporated into anti-racist trainings in business, academia, criminal justice, and other fields. But, while plenty of people accept that bias is a part of the human condition (babies as young as three months show preferences for faces of their own race over others), the idea that this implicit bias leads to specific action (say, calling the police on two black men who sitting in Starbucks) is anything but proven.

The problem, Singal wrote in another post about the test, "is that there’s very little evidence to support that claim that the IAT meaningfully predicts anything. In fact, the test is riddled with statistical problems—problems severe enough that it’s fair to ask whether it is effectively 'misdiagnosing' the millions of people who have taken it, the vast majority of whom are likely unaware of its very serious shortcomings. There’s now solid research published in a top journal strongly suggesting the test cannot even meaningfully predict individual behavior. And if the test can’t predict individual behavior, it’s unclear exactly what it does do or why it should be the center of so many conversations and programs geared at fighting racism."

That, of course, hasn't stopped it from being widely held up as a valuable part of anti-racist training. I asked Singal, who is currently working on a book about fads in social science that includes a chapter on the IAT, what he thinks of Starbucks' plan, and he said that while the company hasn't provided enough information to gauge exactly what Starbucks will be doing with its employees on May 29, "there are reasons to be skeptical.... The IAT is riddled with problems and doesn't even appear to really measure anything at the level of individual test-taker. Plus, even if one assumes the IAT is useful, there's zero evidence anyone has come up with an intervention that can reduce discriminatory behavior by reducing implicit bias." And, as Singal and others have noted, the creators of the test have repeatedly tried to smear those who question its effectiveness.

Starbucks clearly seems to have a problem with race—although, whether it is systemic or not will take a thorough investigation to find out. When I asked the company if they will be implementing the IAT, they did not say, but a spokesperson said, "the curriculum will be developed with guidance from several national and local experts confronting racial bias, including Bryan Stevenson, founder and executive director of the Equal Justice Initiative; Sherrilyn Ifill, president and director-counsel of the NAACP Legal Defense and Education Fund; Heather McGhee, president of Demos; former U.S. Attorney General Eric Holder; and Jonathan Greenblatt, ceo of the Anti-Defamation League."

It's quite a roster, but if the IAT is central to their mission, there are reasons to fear that it just won't work.

Update: Anthony Greenwald, one of the developers of the IAT, responded to my request for comment after this was published yesterday: "Starbucks would be wise to check out the scientific evidence on implicit bias training," he said. "It appears to be the right thing to do, but this training has not been shown to be effective, and it can even be counterproductive. It will appear that Starbucks is doing the right thing, but the training is not likely to change anything. The Implicit Association Test is a valuablle educational device to allow people to discover their own implicit biases. However, taking the IAT to discover one’s own implicit biases does nothing to remove or reduce those implicit biases. Desire to act free of implicit bias is not sufficient to enable action free of implicit bias."