I’ve been taking a look at our data regarding diversity. What I’m seeing demands action.
At Eightfold.ai, we’re a data company. We have anonymized information for more than 1 billion career profiles worldwide, including more than 90 million careers in the U.S. It’s a substantial sample of the global workforce, and our AI adds additional context to each profile, including the skills each person has and the seniority level of each job that they’ve held.
So I wanted to understand how the U.S. workforce is doing on gender diversity. Are we starting to see parity in the workforce by gender? As you see in the chart, we are not doing well enough. The chart shows the share of the workforce by gender at each of seven standard seniority levels. As you look higher up the career ladder, the share of female individuals with positions at each seniority level steadily drops.
By the time you reach the Vice President level, female individuals are only half as likely as non-female individuals to be VP. At the C-level, female individuals are just one-third as likely to have attained the C-suite.
I’ve read how the number of female CEOs in the Fortune 500 (like GM’s Mary Barra, pictured) is on the rise, but it’s still just 37. I wasn’t expecting to find good news in this analysis, but these stark numbers are distressing nevertheless.
There’s a crying need for greater diversity in our organizations, and for greater inclusion to create fair representation at all levels. And not just on the dimension of gender, but on all aspects of diversity that enrich our workplaces and improve the performance of our companies.
Fortunately, we can start to move the needle with technology. The right tools, combined with organizational commitment and intention, can make an impact.
Below I propose four strategies that all organizations can pursue today to finally make an impact.
Strategy #1: Create a Candidate-Focused Career Site
The candidate experience, from the very first moment, harms underrepresented groups. Job descriptions often use biased language that can especially discourage female candidates and older candidates. The simple fact that applying can take so much time hurts busy parents and older workers.
Organizations can fix these problems by making their career site focus on the needs of the candidates first. With technology, career sites can encourage rather than discourage. And if every candidate is encouraged to apply for the job that best fits their potential, that experience will overcome much of the self-selection bias we see.
When your candidate comes to the career site, they should see information personalized for them. The most important question any career site visitor has is, what jobs are available for me? Candidates want to find the best jobs for themselves as fast as possible.
With a candidate-focused career site, all of your available jobs are ranked for the individual candidate, so they can see first the jobs that they are a match for. And they can see exactly why they are a match: what skills and experiences they have that make that job a compelling option for them, and that will make them a strong candidate for your company.
This experience is encouraging because it shows each person why they are a great candidate for a job. This reduces self-selection bias, because different groups of candidates may have different risk tolerance in applying for jobs. Encouragement reduces this sense of risk.
Furthermore, organizations should make applying as fast as possible, so busy parents and older workers can be included.
Making it easy to apply is often not the approach organizations have taken in the past. Some organizations want candidates to prove their interest by jumping through hoops: fill out a long form, take an assessment test, provide a unique cover letter, and so on. But this logic is backwards. This approach drives off diverse talent, in particular, and other strong performers who will decide applying isn’t worth their time.
If you’re worried about having too many applications to screen, note that AI can help you automatically screen any number of candidates today, so this concern goes away entirely.
Strategy #2: Use Anonymous Screening
The hiring process is prone to bias. Recruiters and hiring managers see a candidate’s personal characteristics, such as their perceived gender, ethnicity, age, and educational credentials, and make selection decisions based on these factors rather than on each person’s potential to succeed. The result is reduced diversity.
I’m not suggesting that recruiters and hiring managers are doing this on purpose. Quite the opposite, these biases are almost always unintentional. It’s just very hard for all of us, as human beings, to separate out all the factors that go into our decision making, even if we know intellectually that some factors shouldn’t matter.
So that’s why we need anonymous screening.
With anonymous screening, the hiring manager (and the recruiter, if desired) doesn’t see any personal information about the candidate. They see only skills and qualifications, and no evidence of their gender, age, race, or educational qualifications. The image gives an illustration of how a profile appears before and after it is anonymized.
There is plenty of evidence that anonymous screening works to prevent bias, leading to greater rates of hiring of diverse talent.
Anonymous screening has gone by multiple names. This strategy has also commonly been called candidate masking or blind screening. While these terms all mean the same thing, I think anonymous screening is the more inclusive term.
You might be thinking that anonymous screening can only help so much. After all, won’t the hiring manager eventually find out who the candidate is, and have the ability to judge their personal characteristics? Yes, that’s true—even with anonymous screening, there are still opportunities for bias. Which brings us to Strategy #3.
Strategy #3: Measure and Correct in Real Time
Since anonymous screening can’t address all opportunities for bias, we need to pair it with another strategy. The way to find remaining sources of bias is to use diversity analytics. Not only do diversity analytics find biases in hiring, they can also be used to measure the impact of equity policies.
Diversity analytics show the hiring funnel for each stage, and for each diversity category. The categories can consist of U.S. Equal Opportunity categories, or others.
For each category and at each stage, the analytics show if a statistically significant bias is detected. Let me explain this. Suppose, for example, that 10 percent of all applicants are members of an underrepresented group. This suggests that 10 percent of all hires should be also, but that’s approximate. If 9 percent of hires are members of this group, that might be due to chance. If it’s only 5 percent however, perhaps there is a problem, and the analytics would flag that 5 percent is very different from 10 percent.
In other words, the analytics will show if the outcome is likely different from what would occur due to random chance. It’s then up to you to investigate where and why this discrepancy is occurring.
It’s still possible that the disparate impact is a coincidence—but probably there is a cause. Maybe there is a step in the hiring process that turns away a specific group of candidates. For example, an assessment process may not be accessible and adversely impacts candidates living with disabilities.
In many cases, unfortunately, the cause will be a person who is making biased decisions. Hopefully they just need awareness and training, but you may need to remove the person from a hiring role.
There’s another side of diversity analytics, which is supporting equity policies. I gave the example of a company having 10 percent of its applicants from an underrepresented group and assumed that means approximately 10 percent of hires will identify as members of this group, if the hiring process is unbiased.
But what if the broader goal of preventing bias demands a different target? Perhaps this group represents 20 percent of the community, and individuals identifying with this group have faced historical disadvantages such that they are less likely to apply for a job at the company. Shouldn’t fairness mean that the company makes 20 percent of its hires from members of this group?
In this situation, the company may choose to pursue an equity policy to actively increase the share of applications from individuals identifying with this group. A policy like this may be called affirmative action.
Now, I can’t tell you what affirmative action policies are right for your company. This is a very complex and serious topic area, and it is way beyond the advice I could give you in a blog. You have to consider your local laws, the history of your organization and your society, the expectations of your workforce and your customers, and more. I can’t tell you what you should do.
What I will tell you is that whatever you do, you need to be able to measure the impact. You need to know if the equity policy is having the intended effect or not. And that’s another important aspect of these analytics. They will track the effect of equity policies and reveal whether or not they are doing their job. Measurement is the key to fairness.
I want to compare the importance of measuring equity policies to a subject we are all becoming familiar with: vaccine clinical trials. A vaccine must be carefully tested for safety and for efficacy in clinical trials. If you have a new vaccine, just because you think it’s going to work based on a computer analysis, just because you have a good and logical argument, that’s not nearly enough to get people to trust it.
You need the stats. You need to prove that the vaccine does what it’s supposed to do, and nothing serious that it’s not supposed to do. Equity policies need to be tested in a similar way.
With diversity analytics, you can find and remove the biases that still exist, and lay the foundation for greater fairness in your company.
Strategy #4: Offer Skills-Based, Self-Service, Transparent Talent Management to All Employees with AI
The fourth strategy is, in many ways, a culmination of the first three applied to the talent-management services you offer to your employees. By basing these services on the individual skills of each employee, by making them self-service, by offering transparency, and by creating one policy for all employees, you can create lasting change.
Talent management services can include several components: learning & development, mentorship, job transfer & promotion, projects & gigs, and more. You may define the services that you offer your employees differently. Regardless of how you structure talent management, placing an AI platform for talent management behind your employee career services will create an inclusive basis for those services.
Here’s why this change is so important:
- With a skills-based approach to talent management, the standard for career advancement becomes what each employee can do, and not who they are. And, a skills-based approach is encouraging to all employees.
- Self-service for talent management lets each employee explore and decide on their own time, without the pressure of a performance review or HR meeting.
- Both skills and self-service create a transparent career experience. Every employee finds what options are available, and knows that the same options are available to others. Each employee sees how the company will make decisions. When employees can understand how a decision such as promotion will be made, they will gain confidence in the fairness of the process.
- Finally, this platform is for all employees. Employees with social advantages can’t work around it. Mentors can’t choose who to coach on the basis of their biases. Important projects can’t be staffed at happy hour.
With the right technology supporting talent management, your company can change the entire mindset around careers. Employees will see the presence of opportunity, and see fair outcomes in action. In time, the expectation of facing bias will be replaced by an expectation of full participation. When this change in mindset is achieved, I believe, we will finally have real inclusion.
Conclusion: The Right Career for Everyone in the World
The strategies I’ve shared above speak to our motivation at Eightfold.ai. We believe that technology can create a triple benefit—to companies, to individuals, and to society.
We express this motivation with our mission: “The Right Career for Everyone in the World.”
The right career is the career that each individual chooses to best express their interests and their talents. Just imagine the scale of productivity, motivation, and personal joy that would be unlocked if every person could truly pursue their right career. How many of our problems in society could we solve?
Everyone in the world is an inclusive group. Everyone includes people in their first job and people looking toward retirement. Everyone includes every race and ethnicity, and every religion and creed. Everyone includes people of all genders. Everyone means everyone.
And so, the right career for everyone is only possible if diversity, equity, and inclusion are a core part of each company’s objectives. The work ahead of us isn’t easy, but we can make a real difference and move forward toward our goal. We can advance the day when our data will show an equitable workforce in the United States, and everywhere else. Let’s start now.