The call for greater diversity in the boardroom and beyond hasn’t yet yielded significant change. Most efforts progress by inches, but companies that take a new tack to address unconscious bias and build a more inclusive workforce could turn the tide on gender issues. In this interview with McKinsey’s Rik Kirkland, Harvard Kennedy School professor of public policy Iris Bohnet talks about what is working—and what is not—when it comes to building a more equitable workplace. An edited version of her remarks follows.
[There’s] great work showing that diversity is correlated with business performance. But we’ve also learned that that probably won’t be enough to move the needle. That came as somewhat of a surprise to many of us, who thought that if we can show the business case, things will happen. But clearly more needs to be done.
The failures of diversity training
About $8 billion a year is spent on diversity trainings in the United States alone. Now, I tried very hard to find any evidence I could. I looked not just in the United States but also in Rwanda and other post-conflict countries, where reconciliation is often built on the kind of diversity trainings that we do in our companies, to see how this is working.
Sadly enough, I did not find a single study that found that diversity training in fact leads to more diversity. Now, that’s disappointing, discouraging, but maybe when we unpack it also understandable. The unpacking means that there’s a lot of research that has nothing to do with diversity or gender or biases but is more generally trying to understand how people think, and it has shown that it is actually very hard to change mind-sets.
Based on that evidence, maybe we shouldn’t be quite as shocked that diversity training doesn’t have the impact that we were hoping it could have. Because even though you and I might agree now that we will be inclusive tomorrow, it is hard to follow through on those virtuous intentions.
How design can address unconscious bias
What we’re up against often is referred to as unconscious bias. It means that if I think kindergarten teacher, I don’t think man. And if I think engineer, I don’t think woman. Seeing really is believing.
A powerful study demonstrating unconscious bias was actually run with orchestras. In the 1970s, some major US orchestras introduced blind auditions. They had musicians audition behind a curtain and then evaluated their performance. The interesting thing about this design feature, this curtain, is that it was introduced despite the fact that many of the orchestra directors thought that they of course didn’t need curtains—that they of all people only cared about the quality of the music and not what somebody looked like.
It turns out that curtains helped increase the fraction [of women] on these orchestras from about 5 percent in the 1970s to almost 40 percent now. That is the power of design. The curtain is important for me for two reasons. On the one hand, it is a real example showing the power of unconscious bias. But it is also important because it helps us understand that sometimes we have to make it easier for well-meaning people to do the right thing.
What works to promote gender equality
There are things that do work if you design them right. Talking about talent management, for example, we can go through the life of a person once he or she enters an organization. That starts with sourcing talent.
Most organizations would argue that they’d like to benefit from 100 percent of the talent pool. One way to do this is to start with our job advertisements and de-bias the language that we use in them. We have to scrutinize the kinds of descriptions that we use in our job ads. Let’s cast the net widely and use language that’s inclusive. Not every word can be expressed as a gender-neutral word. But what the research suggests is that if you use a very gendered word like assertive, which may be an important characteristic you want to look for, counterbalance it with a word such as cooperative.
We enter more difficult territory as we start to evaluate people. It’s difficult because most of us believe we are very good at it, when in fact the evidence suggests that’s not true. We are very likely to be influenced by what somebody looks like, when that’s actually not a good predictor of performance. We’re influenced by whether we share the same hobby, have the same accent, come from the same country—lots of things that in theory shouldn’t matter.
What can you do there? At the first stage of the evaluation process, I would recommend that companies blind themselves to the demographic characteristics of job applicants. That means taking off names. In some countries—such as Germany or Switzerland—age is still included [on applications]. In many countries, you still add a photograph to your job application. All of that should go.
Here we can really learn from the orchestras and try to help our minds focus on the quality of the candidate, not whether somebody looks the part. One example: a start-up, Applied, tested the impact of blinding. It worked with a tech company and had every applicant go through the traditional process. In parallel, every application was reviewed using the Applied process, which included blind evaluations. What this tech company found in the end wasn’t so much gender, racial, or other biases, but rather disciplinary bias. It had thought it was only looking for computer scientists and engineers—a small sample of the general population. Once the company blinded itself to some of those characteristics and relied on job sample tests where people were confronted with some of the tasks they would actually have to do, it started hiring neuroscientists, psychologists, people who could do the work but wouldn’t naturally fall into the category it would hire from.
Changing the default to drive change
One of the early insights in behavioral science was that defaults matter. It really matters where we start our assumptions. We just heard from a company that changed the default in their job ads to part-time work, saying that the default is part time, but you can opt out and work full time if you’d like to.
Telstra is a big telecommunications company in Australia. It changed the default to flexibility. Every job ad now says that flexible work is the assumption. And in its firm culture, the norm is basically to ask, “Why are you in the office today? Couldn’t you work from home?” Already, from Telstra’s data, I know that it increased the likelihood that women would apply dramatically.
The next horizon: People analytics
I’m not arguing that we should leave this up to machines. But I am saying that we should use machines, algorithms, and data much more intelligently together with humans in making those decisions.
This is an important insight. We’ve been throwing money at the problem through diversity-training programs and leadership-training programs, trying to help traditionally disadvantaged groups, including women but also people of color and people with disabilities. That is not the way to go.
We have to understand what’s broken and then intervene where the issues are—really tease apart what’s broken, and then try to fix it and use data on what works to inform our decision making. I am quite optimistic that big data analytics and experimentation will move the needle dramatically in the next ten years. But I am mentioning experimentation also to suggest that we don’t have all the answers yet.