What the “Bias of Crowds” Phenomenon Means for Corporate Diversity Efforts
Originally published on behavioralscientist.org.
By Liz Kofman-Burns
Despite widespread recognition that diversity in the workplace yields tangible benefits, despite viral stories exposing hostile company cultures, despite millions (if not billions) of dollars spent on unconscious-bias trainings, the long-promised diversity revolution has stalled. Why are organizations failing to make good on their diversity efforts?
The answer may be found in another puzzle, one that has perplexed psychologists in the field of implicit bias. The puzzle goes like this: Several recent studies have shown that individual biases are unstable and are only weakly associated with actual behaviors. Studies also show, however, that implicit bias at the aggregate level (e.g., nations, states, metropolitan areas) is stable and reliably predicts discriminatory behavior. (For example, in regions of the U.S. where implicit race-bias is higher, police shoot disproportionately more black citizens.)
Further complicating the implicit bias story, a widely used tool to measure an individual’s implicit bias, the Implicit Association Test (IAT), has come under fire in recent years for methodological inconsistencies. This doesn’t suggest that implicit bias doesn’t exist, but rather that it can be difficult to measure and quantify consistently.
On one hand, differences in implicit bias exist and are greatly consequential; on the other hand, they’re difficult to measure, and to link to individual actions.
From a company perspective, if implicit bias is unstable and doesn’t predict behavior at the individual level, are company-sponsored efforts to mitigate bias unfounded?
The short answer is no. Companies can mitigate biased behavior, but first they have to accept a paradigm shift in how implicit bias works.
The “bias of crowds” model of implicit bias
In a recent study in Psychological Inquiry, researchers at the University of North Carolina and the University of Richmond present a new theory that helps solve the implicit bias puzzle and provides a blueprint for how companies can make a real impact on workplace diversity. They call it the “bias of crowds” phenomenon. They argue that implicit bias should be understood first and foremost at the situational level, not the individual level. That is, bias is based on malleable social categories, thoughts, evaluations, and stereotypes. In certain situations, individuals have greater access to biased concepts. In other situations, these biased concepts are harder to retrieve.
In practice, this means that if you work in an environment where biases and stereotypes are easily accessible, bias against certain groups is going to be more commonplace. Consider the example of GoDaddy, an internet domain registrar. For several years, GoDaddy’s advertising strategy centered on selling website hosting via images of scantily clad women. Staff openly complained that the offensive advertising corroded their workplace environment. GoDaddy eventually faced a sexual harassment lawsuit and had trouble attracting talented engineers and executives.
Of course, an environment where bias flourishes can be far subtler than one where overtly sexist ads are promoted. If a recruiter inherits job descriptions that say the company is looking for “rock stars,” “code ninjas,” and engineers who can “wrestle problems to the ground,” the recruiter may understandably, even unconsciously, move more resumes from men than women to the top of the pile. The same recruiter’s actions might be completely different if the company’s job descriptions have a less aggressive tone and instead stress qualities like “collaboration,” “sensitivity” to client needs, and the ability to develop “warm relationships.” The recruiter is still the same person, but the context they are working in—the environment the company has fostered—will result in the recruiter taking different actions.
The key takeaway from the bias-of-crowds model is that bias often reveals more about the workplace environment than about the individual. Practically speaking, it means that instead of focusing on individuals, organizations that want to improve diversity must focus on structural change.
Yet, for years now, companies attempting to mitigate bias have largely focused on the opposite: individual level interventions. Influential companies like Google have touted the benefits unconscious bias training for employees; because Google does it, many other companies have followed. But the vast majority of companies, including Google, have little to show for these individual-centric attempts to reduce bias.
Here are three recommendations for how Google, and other companies, can leverage the bias-of-crowds model to make good on those diversity promises.
#1 Focus on changing processes, not people
Let’s return to the GoDaddy example. When faced with a sexual harassment lawsuit, its leaders didn’t simply offer sexual harassment training to managers and call it a day. Company leadership analyzed all aspects of their human resources processes for systematic biases, including in hiring, performance reviews, and promotions. When they found systematic problems, they changed the systems. For example, they changed employee evaluation forms to better assess employee impact rather than rely open-ended questions. As a result, GoDaddy saw promotions of women jump 30 percent in a single year.
The key to changing a process is to scale structured processes. In my work at Talent Sonar, I have seen companies make great strides in changing their hiring processes without even mentioning the word “bias” (a tactic that responds to a documented backlash effect).
For example, companies that want to systematically attract more candidates from underrepresented backgrounds empower their recruiters and hiring managers to consistently screen job descriptions for inclusivity. Some companies understand diversity not as a pipeline problem but rather as a leaky funnel problem, and they have changed their résumé review screening processes. Instead of having hiring managers spend an average of just 6 seconds scanning a candidate’s name, school, and previous employer (none of which are good predictors of job success), evaluators at these companies review résumés with candidate names and school names redacted, pre-commit to which categories are most important, and score them for each candidate. Likewise, companies that care about reducing the many biases that occur during interviews have shifted to using structured interviews, which have been shown to greatly mitigate bias regardless of whether a candidate shares an interviewer’s alma mater or love of the Red Sox.
#2 Prioritize process change and stick to it
It’s important to understand that process change is difficult. When Uber was accused of pervasive gender discrimination, an outside investigator produced a report recommending sweeping structural changes to the company—everything from new reporting mechanisms to blind résumé review to structured interviews. Uber’s board of directors publicly accepted these recommendations. The challenge for Uber, and for many companies, will now be whether there is enough organizational will and discipline to implement and sustain these new processes.
In my view, too many companies are searching for a “magic bullet” that will allow them to meet their diversity goals without the hard work that process change requires. It’s easier to assume that the problem lies in a few bad actors, rather than with company policies and practices. It’s easier to host an unconscious bias training, even an expensive one, than to systematically change how everyone in your company conducts interviews.
Even when companies want to try new processes, a problematic pattern often emerges: Companies pilot dozens of programs but don’t collect the right data to measure results. Their employees get pilot fatigue. Buy-in dissipates, morale declines, and resentment sets in.
Instead, company leadership should prioritize a few process changes and provide the resources and time necessary to carry them out.
#3 Provide resources and incentives for change management
Implementing new, structured processes is a major challenge that requires resources and time. For example, changing hiring practices from ad hoc to structured will usually require hiring teams to slow down, at least at first. Yet recruiters and HR departments are often incentivized based on “time-to-fill.” That is, the faster they make a hire, the better. In my experience, recruiters are rarely evaluated or given bonuses on the quality of a hire. In addition, hiring managers and interviewers often see hiring as outside of their “real” jobs. Here, the onus must be on company leadership to create a context where fairness, intentionality, and bias-mitigating structures are built-in.
The shift toward strategic process change in HR is already underway. The companies that understand how much their everyday processes influence the bias of crowds in their workplace and start the hard work of tackling process change today will be the winners of the talent wars tomorrow.