From Google to Starbucks, more companies are introducing implicit bias training. These HR-sponsored courses are intended to foster diversity and inclusion by making employees more aware of unconsciously believed negative stereotypes. The idea is that if we can combat our underlying biases, we’ll decrease discriminatory behaviors at work and level the playing field for women and underrepresented minorities.

And yet despite the growing adoption of unconscious bias training, there is no convincing scientific evidence that it works. In fact, much of the academic evidence on implicit bias interventions highlights their weakness as a method for boosting diversity and inclusion. Instructions to suppress stereotypes often have the opposite effect.

This is because the main problem with stereotypes is not that people are unaware of them, but that they agree with them. In other words, most people have conscious biases. For instance, in virtually any culture men are more likely to believe that women are too kind and caring to be leaders. From an early age, our relationships with others are shaped by common cultural stereotypes about social class and status. Perhaps most obviously, every nation has pervasive cultural stereotypes about other nations, usually its neighbors, which play a critical role in shaping and cementing their cultural identity.

Moreover, to the extent that people have unconscious biases, there is no clear-cut way to measure them. The main tool for measuring unconscious bias, the Implicit Association Test (IAT), has been in use for 20 years but is highly contested. In simple terms, the test measures people’s reaction time in response to different categories of words or photographs.

The idea is appealing — ever since Freud, people have been seduced by the notion that unconscious motives are the secret causes determining our behavior. But meta-analytic reviews have concluded that IAT scores — in other words, unconscious biases — are very weak predictors of actual behavior. In other words, the vast majority of people labeled “racist” by these tests behave the same as the vast majority of people labeled “non-racist.” Do we really want to tell people who behave in non-racist ways that they are unconsciously racists, or, conversely, tell people who behave in racist ways that they aren’t, deep down, racists at all?

This gets to the underlying flaw with unconscious bias trainings: behaviors, not thoughts, should be the target of diversity and inclusion interventions.

It’s tempting to think that simply making people more aware of their biases will automatically level the playing field. But the reality is more complicated. Scientific evidence suggests that the relationship between attitudes and behaviors is much weaker than one might expect. People often believe in the benefits of corporate diversity yet fail to act on those good intentions — just as those who support organ donation don’t always choose to donate their organs, and those who believe in recycling don’t always recycle.

In fact, in most work-related settings the majority of people at least occasionally behave in ways that run counter to their attitudes — whether that means accommodating a micromanaging boss or agreeing to take on a dull project. “Faking good” is a key part of good organizational citizenship.

Even if we lived in a world in which humans always acted in accordance with their beliefs, there would remain better ways to promote diversity than by policing people’s thoughts.

Organizations should focus less on extinguishing their employees’ unconscious thoughts, and more on nurturing ethical, benevolent and inclusive behaviors. This means focusing less on employees’ attitudes, and more on organizational policies and systems, as these play the key role creating the conditions that entice employees (and leaders) to behave in more or less inclusive ways. Instead of worrying what people think of something or someone deep down, we should focus on ways to eliminate the toxic or prejudiced behaviors we can see. That alone will drive a great deal of progress.

Tomas Chamorro-Premuzic is chief talent scientist at Manpower Group and a professor at Columbia University.

In a time of both misinformation and too much information, quality journalism is more crucial than ever.
By subscribing, you can help us get the story right.