Understanding the science of hiring bias: webinar follow-up

Lauren Shufran

Lauren Shufran

Content Strategist

Posted on

March 3, 2021

Last week, Gem partnered with Greenhouse to offer a webinar called “Understanding the Science of Hiring Bias (and How to Overcome It).” Our CEO and Co-founder, Steve Bartel, joined Gary Davis (Diversity, Equity, & Inclusion Advisory Director at Greenhouse), Riham Satti (Co-founder and CEO of MeVitae), and Erin Thomas (VP and Head of both Talent Acquisition and Diversity, Inclusion, & Belonging at Upwork) to discuss everything from how bias can manifest during the hiring process and how to mitigate it, to how to source more diverse candidates by reducing reliance on biased qualifiers, to why data is so crucial to diversity hiring efforts, to what technologies can help teams hire most fairly and effectively. We received a ton of great questions from folks both prior to and during the webinar, and we couldn’t cover them all in the time we had. So we did what we could to cover them below. If you missed the webinar and want to catch the great insights that were offered, you can find a recording of it here. If you attended but your question wasn’t covered, we hope you’ll find it below:

Is there any correlation to how a job description is written and attracting a diverse candidate pool? If yes, any tips? / What language should we be using/avoiding in our job ads in order to attract more diverse candidates?

There’s absolutely a correlation between how your JD is written and whom you’ll attract or turn off based on the language found there. From the point of view of diversity, “mature” or inclusive job descriptions communicate clear information about what's expected of the applicant and what they can expect from the company. They’re free of jargon and devoid of excessive requirements. They provide key insights into organizational culture and are free of language that may signal to underrepresented talent that they won’t feel safe or welcomed in the workplace. They reduce requirements to “must-haves,” or are results-based rather than requirements based. If you want more details about what each of those things means, we’ve written more about inclusive job descriptions here.

How do we best advertise in our job descriptions that we’re an inclusive company, and that all applications will be analysed objectively?

If you employ the best practices we discuss in the post we linked to above, talent may intuit the care you’re taking to build an inclusive workforce. But it also doesn’t hurt to just say it. Instead of claiming you’ll accept applications from all demographics, specifically encourage talent from underrepresented demographics to apply. “We strongly encourage people from underrepresented groups to apply” is one of the more common ways of phrasing this inclusive strategy. Just note that the data shows inclusive job descriptions with equal opportunity language beyond the boilerplate statement fill, on average, 10% faster across all demographic groups than descriptions that don’t include such language. Which means a statement about your company's commitment to DEI has to be customized, genuine, and human, not cut-and-paste and superficial. Prospective candidates will believe it only if it aligns with the other language cues in your inclusive job description.

When is the best stage in the hiring process to ask about preferred pronouns? We would like to ensure people feel as welcome as possible, and so would like this information pre-interview if possible, but also want to avoid any fear of bias for the applicant. / Any strategies on collecting identity data at various funnel stages, given the hurdle of this submission being *optional* for candidates?

At Gem, we recommend asking as soon you can. Some of our customers have been experimenting with a prospect self-identification form, and it’s an option worth considering. A prospect self-identification form is exactly what it sounds like: an opportunity for prospective candidates to self-identify before they apply—and before you can officially gather EEOC data. One option is to include a link to the form in your signature for all prospect outreach. Another is to actively ask prospective candidates to consider filling it out once they’ve replied as “interested.”

There are any number of ways to phrase the messaging around this. One up-front way is to say something like: "At [company], we know how important it is to have a diverse team in order to have a strong one. That's why we want to build relationships with as broad a range of talent as we can... even before you apply. If you’d like to help us do that, we’d love for you to fill out this short form to tell us a little bit about yourself." Another option which doesn’t so explicitly refer to your diversity initiatives might read something like: "I'm so glad to hear you're interested in hearing more about us! At [company], we're serious about starting relationships off on the right foot; and we want to make sure that we call you what you want to be called, refer to you the way you want to be referred, and send you the content about our company that's most relevant to you. If you have 2 minutes to tell me a bit about your preferences before we chat, it'll help me get to know you... before I get to know you."

However you decide to phrase it, the point, of course, is to underscore that diversity and conscientiousness around the various ways talent identifies are both values you prize. Make sure you note that talent’s responses will not be linked to any application or be used to influence hiring decisions. For organizations that de-couple demographic information from applications as a bias-management strategy, this is key.

If there is a strong company culture within an organization and hires are made who will fit in with company culture, is this considered a biased hire? How to justify a hire based off of culture fit within the organization without a bias?

We’ve written about this elsewhere, too, and it’ll do your team a world of good to figure out how to think about “culture add” and “values fit” as much as (or more than) you think about “culture fit.” A lot of companies are dispensing with the term “culture fit” right now, and maybe for good reason: the term leads people to focus on how a candidate will adapt and assimilate into a preexisting culture. People tend to think that “alignment” means sameness—something a 2012 study called “Hiring as Cultural Matching” found. The study found that “shared experiences” was one of the most important factors used to determine culture fit. “Alignment” translated to “similarity of backgrounds”—and therefore, “similarity of viewpoints”—for most hiring managers. Over time, the term has taken on a tribal meaning, facilitating bias by giving recruiters, hiring managers, and companies permission to hire people that walk, talk, look, and think like them.

In recent years, some alternatives have arisen among companies and talent teams that reject what “culture fit” has come to stand for.

The first is “culture add,” and it’s exactly what it sounds like: “The likelihood that someone will not only reflect the company’s values and professional ethics, but also bring diverse opinions, experiences, and specialized skills which enhance not just the team, but the overall company culture.” In this definition, culture add takes the best aspects of “culture fit” (shared goals, for instance), advocating for the behaviors that will lead to organizational success. But it adds an important element: Where are the company’s cultural blindspots? What perspectives, experiences, and approaches are we missing? Where might we need our thinking and our processes challenged, and what kind of person would challenge them? You get both values-alignment and value-add here; and because you’re seeking experiences and perspectives that you’re currently lacking, homogeneity will never be a struggle. 

The second is “values fit.” Aubrey Blanche, currently Director of Equitable Design, Product, & People at Culture Amp, describes values-fit as “hir[ing] people who share our goals, not necessarily our viewpoints or backgrounds.” Our recommendation would be to find ways to incorporate one, or both, of these alternatives into your hiring decisions.

What are some tangible places to source candidates of color beyond LinkedIn? Any platforms that you've been pleasantly surprised by? / What job boards do you recommend using to source more diverse candidates? / On LinkedIn, where can I find a list of the groups Steve mentioned that we can use to target BIPOC candidates? 

We’ve got answers to all of these questions in Diversity Sourcing at the Top of the Funnel: Where to Find Underrepresented Talent

As an organization, we do a pretty decent job of diversifying talent pools for entry-to-senior level positions, but struggle more when it comes to leadership roles. Is there any advice you can provide specific to VP and C-level roles?

There are any number of ways to approach diversity at the leadership level in your org. One that we love is from our customer SurveyMonkey, who has implemented a “diversity sprint” for director-and-above roles. For these roles, the team sources and screens only underrepresented talent for 30 days before turning to applicants or referrals. (You can hear Becky Mohr, Director of Talent Acquisition at SurveyMonkey, discuss that strategy here.) At Gem, one of our strategies is to have our own leadership team increase the diversity in their professional networks. This way, when they make referrals for their roles, those referrals will be more diverse. Leadership is committed to making time for three coffee chats with underrepresented talent every quarter. It’s a longer-term strategy, for sure; but we want to make sure that we’re nurturing relationships with great talent for when those leadership roles open—even if it’s a year or two down the road. 

How do you measure success in diversity hiring? Are there any benchmarking tools out there to "grade" ourselves on how we compare to other employers in similar industries and understand what other employers may be experiencing?

At Gem, we have our own benchmarking report, where you can find passthrough rates by gender—from outreach all the way through offer-accept—to benchmark yourself against your peers when it comes to gender equity. Thanks to our new race/ethnicity feature, we’ll be able to offer passthrough rates for underrepresented minorities in future reports. We’d also suggest you take a look at orgs of similar sizes in your industry to see what kinds of goals and numbers they’re publishing. But in some cases, using internal comparisons to understand your team’s performance— gauging success by percentage of improvement over last quarter or year, for example—is really insightful, especially if you’re just getting started on your diversity strategy. 

How can we ethically ensure that we have the necessary demographic data at the top-of-funnel stage (especially at the sourcing stage) without, or before, knowing how a prospective candidate identifies? / What are some strategies or ways for organizations to start that data tracking process of their pipeline? / How to collect metrics?

Above, we discussed self-identification forms, which will always be the best way to get demographic data from your prospects and candidates since it’s directly from the source. But we also like to think that Gem is the answer to all of these questions. Two of the features built into our product are a gender feature and a race/ethnicity feature that track certain demographics (male, female, nonbinary, Asian, Black, Hispanic/Latino, White, and undetermined)—in aggregate—through the funnel. Predictions are determined by a candidate’s first and last name, as well as by location for race/ethnicity, via a model trained on large datasets. In aggregate, gender results are 95% accurate, and race/ethnicity results are 75%-95% accurate. 

Because this information is for directional guidance on pipelines rather than for tracking candidates on an individual basis, race/ethnicity predictions are viewable only in aggregate in both our Outreach Stats and Pipeline Analytics. The prediction is not displayed on candidate profiles, and aggregate stats are not visible if the sample size is too small. Ultimately, these demographic insights allow talent teams to track how their work affects diversity through the recruiting funnel, from first outreach through conversion to hire. You can use these insights to:

  • Catch whether certain groups are disproportionately dropping out of the funnel at a certain stage

  • Forecast how many reach-outs are needed to convert one underrepresented hire

  • Understand via top of funnel metrics why the team did or did not hit diversity hiring goals

  • Inform hiring managers about diversity breakdowns in historical data

  • View differences in pipelines between focused effort on diversity and sourcing through unfocused methods

  • See whether messaging is resonating more or less with a specific population

I would love to hear any suggestions you might have for making an employee referral program more diverse.

We all know that relying on referrals can complicate diversity initiatives. Because employees are more likely to refer talent who is demographically similar to themselves, referrals tend to benefit White men more than men of color or women of any race. Indeed, according to a report from PayScale, White women are 12% less likely, men of color are 26% less likely, and women of color are 35% less likely to receive a referral than White men are. That means a lot of majority talent is organically coming your way.

The alternative is to take a more active approach to employee referrals and put strategies in place that prioritize underrepresented talent. One strategy is to leverage your employee resource groups (ERGs) or minority employees. If our social and professional networks tend to be made up of people who are demographically similar to us, then the best sources for diverse referrals will be these employees. Leveraging the “similarity effect” with underrepresented employees lets you see all the benefits of referral hiring while simultaneously diversifying your workforce. You can also search the connections of minority employees at your organization. These are connections employees might not think to refer. (Employees often refer people they know are looking for work, rather than the best person they know who’s doing that job. And you want the best person they know who’s doing that job). When you uncover strong prospects connected with an employee, send that employee a link to their profile and simply ask them if they’d refer that person. If you get a “yes,” you know what to do next.

Another strategy is to simply qualify what you’re looking for in your referral program. Gem’s customer Pinterest discovered one trick to moving the needle on their diversity hiring initiatives by changing the language of their ask. “We found that people tend to refer job candidates who look like themselves,” said Candice Morgan, Head of Diversity at Pinterest, “so we decided to ask people for loose connections and leads instead of referrals.” Some research suggests that another way to diversify is through a referral bonus. Findings from the Kapor Center, for example, suggest that offering an employee bonus for referring URMs moves the needle on diversity initiatives more than other strategies do. We’d advise you to think carefully about whether your company is ready for such a program: If implemented carelessly, it might feed into perceptions that URMs are “special” in an undeserving way, and can contribute to regrettable power dynamics and values systems (“caring” about diversity for money’s sake). 

Finally, consider organizing a source-a-thon or referral-a-thon. These events are exactly what they sound like: social gatherings with your team that involve friendly competition to see who can refer (or source) the most prospects in a given amount of time. (Okay; maybe they also involve pizza.) Kick off the event by educating everyone—from employees to hiring managers— on best practices for searching and sourcing, equip them with templates for their outreach, and let them craft and send the outreach themselves. You can also use these sessions to have them reach out to personal connections they think would be a good fit for your company. Emphasize that the purpose of the event is to source a diverse pool of prospects from a wide variety of backgrounds.

Any thoughts on having a specific % of candidates (ex such as 1/3 of the pool) be diverse or from under-represented backgrounds before moving through each interview stage?

We think that setting these numbers and holding yourself to them is critical. If you watched Becky Mohr’s talk on Operationalizing SurveyMonkey’s Diversity Sourcing Strategy, which we mentioned above, you’d have heard Becky mention that they have a “1-in-4” rule in place at SurveyMonkey that’s been invaluable in helping continue to diversify their team: 25% of candidates who come to onsites for a role have to identify as female, and 25% have to be from an underrepresented minority. A lot of our customers have also implemented the Rooney Rule or the “Rooney Rule +1”: at least one (or, in the case of the Rooney Rule +1, two) underrepresented candidates have to be considered in the slate of candidates for any given role to move forward. If you’re considering it, do it! It will ensure that your sourcers and recruiters are being as thoughtful about the pool as possible. 

How do you intelligently spot unconscious bias? How do you bring unconscious bias to the attention of Hiring Managers in a responsible and respectful way?

Our answer to this at Gem is: data. Numbers make it so much easier to spot bias and to have those conversations, because they’re objective. Take passthrough rates, for example. With Gem’s gender breakdowns, you might observe that many more male candidates than female candidates are passing phone screens, but that female-identified talent that makes it to onsite is getting offers at twice the rate male-identified talent is. This would suggest a bias that’s presenting itself at the top of the funnel for hiring managers, who are letting fewer women into process despite the fact that those women are outperforming their male counterparts once they reach the onsite stage. With the actual passthrough data, you can bring that unconscious bias to hiring managers’ attention (and you don’t even have to call it “bias” if you’re not yet comfortable using that word). Show them how well women are doing once they get to onsites. Explain to them that if they brought in more women at the top of the funnel, they would likely see the same (or better) success rates that they see with male candidates. 

This is just one such example. The point is that the more data you have around the gender, race, ethnicity, etc. of the talent your team is reaching out to, and the talent that is entering and passing through your funnel, the easier it’ll be to spot bias (hint: it’s at the stages certain demographics are leaking out of the funnel), and easier to have conversations around how to mitigate it. 

There were many more questions we weren’t able to get to here; but we also know this is a critical and ongoing conversation. At Gem, we’re not only continuing to iterate on our product for the sake of your diversity sourcing and hiring efforts; we’re also continuing to work on DEI resources to help you with best (or better) practices. So keep an eye out here, and we’ll keep updating you as new content comes out. 


Get started with Gem today

Source faster, build better relationships, and hire more quality candidates.