The nation’s top anti-discrimination-on-the-job officer has a close eye on artificial intelligence

Best listening experience is on Chrome, Firefox or Safari. Subscribe to Federal Drive’s daily audio interviews on Apple Podcasts or PodcastOne.

Whether in the public or private sectors, technology has arrived at the task of recruiting and hiring people. Now, more employers are adding artificial intelligence to their technology stacks. AI has the potential for misuse, or even for introducing unlawful bias all by itself. Charlotte Burrows, chair of the Equal Employment Opportunity Commission, joined the Federal Drive with Tom Temin to share her concerns.

Interview transcript: 

Tom Temin: Miss Burrows, good to have you on.

        Join us Apr. 23 and 24 at 1 p.m. EST for Federal News Network's CX Exchange where we'll be hosting leaders and industry experts to get insight on how agencies are embedding CX into their missions. | Register today!

Charlotte Burrows: Thank you. It is such a pleasure to be here, Tom.

Tom Temin: And the EEOC has launched a specific initiative really to get at the different aspects of the AI issue. Tell us about that initiative.

Charlotte Burrows: Absolutely. The EEOC is going to be taking really a closer look at how technology is fundamentally changing the way that employment decisions are made. Our interest is really in all forms of tech-based employment decision making, including, but certainly not limited to the use of artificial intelligence or AI. This new initiative will build on our past work in the area, and take a close look so that we can help support and guide applicants, employees, employers and technology vendors, in ensuring that these new technologies are used fairly meaning consistently with our Federal Equal Employment Opportunity laws. And that will really allow employers and employees to benefit from these new technologies, while ensuring that they don’t inadvertently become a high tech pathway to discrimination.

Tom Temin: Early on in the internet and online era, a lot of employers introduced machine reading types of things for large numbers of resumes that would come in, and it’s easy to see how the keywords you use and so forth, can, as you say, amplify the discrimination that might be occurring, or even introducing it if they have, you know, lower motives, you might say. Is your sense that artificial intelligence takes this to a new and more complex level, perhaps?

Charlotte Burrows: Well, it is certainly the case that this has been sort of rapidly accelerating. And that on the one hand it’s extremely exciting. But we have to be sure that those civil rights laws that have long protected employees on the job continue to be protected, even as the way hiring is done in this country evolves. So we’ve had, just to make it very concrete, at least one instance that was widely reported in which a tech company tried to create an algorithm to screen potential candidates for software developers, but had to actually abandon that effort after finding that it was systematically downgrading qualified applicants. And so what happened was really a situation where they had taken about a decade’s worth of successful applicants resumes, and basically train the computer to find similar kinds of resumes. But what was happening, because most of those successful applicants had been men, was that if you had the word “women’s” in your resume, so for instance, if you said, hey, I was the captain of the women’s basketball team, that would get you downgraded. Or if you had, in this instance, a couple of women’s colleges on your resume, that would also get you downgraded because the mostly male, previous applicants had not had those characteristics. So it can creep in in ways that are deeply disturbing. And I think most employers would not want and that’s what we’re really trying to both raise awareness about and figure out how to assist in ensuring that that doesn’t happen.

Tom Temin: And you will be launching a series of listening sessions with key stakeholders and talking about some of the tools they’re using. Who do you plan to be speaking with, a range of employers large and small? And also, will you be talking to federal hiring practitioners that might be using these types of tools?

Charlotte Burrows: Absolutely. That’s exactly right. I think just to put this in context, we’ve got about 83% of employers in this country using some form of AI in their hiring. And so we really are going to be talking to employers and also vendors, but really just key experts, so some of the folks who were actually designing these so that we understand how they’re being used. Because as I mentioned in the beginning, this is a rapidly evolving area. And we are also going to be working with other key federal agencies, including OFCCP, and the White House Office of Science and Technology Policy on all of these issues.

Tom Temin: Yeah. OFCCP is the one that oversees federal contractors, and they’ve got some mandates of their own on this front, then.

        Read more: Artificial Intelligence

Charlotte Burrows: That’s absolutely right. They’re a key partner and they have similar authorities with respect to enforcing anti-discrimination protections as the EEOC does, but they look only at those large employers typically, but those employers that are federal contractors, and that’s their authority. So it’s a nice complement and we work frequently with them on a number of different priorities.

Tom Temin: We’re speaking with Charlotte Burrows. She is chair of the Equal Employment Opportunity Commission. And are there any good stories that you know of? The well-known horror stories, you know, are well known because they’re horror stories. But do you have any sense of people that have been maybe using AI to eliminate the bias that might have been in their regular hiring systems?

Charlotte Burrows: Well, that is certainly the goal. And it’s a goal that we applaud, I don’t know that I can identify one particular products, and I’m not sure it would be appropriate for me to sort of advertise that in any event. But that is absolutely one of the reasons why employers are looking at this. There’s the hope that having an automated hiring process can help, in some instances, to eliminate human bias. And so that is, if you will, the holy grail in some of these practices. And what we think is important is that the EEOC help vendors and others understand what our laws require, and the ways in which bias can creep in so that we can assist as they look for that. So I think there’s that potential. I don’t know for certain that that has been fully realized. But I am hopeful that it could be.

Tom Temin: And as you reach out to industry, are you also looking to build up the technical chops of the EEOC staff itself with respect to some of these tools?

Charlotte Burrows: That’s right, and also working, again, with those that have that expertise in the government already. And part of these listening sessions will help us understand really, truly where we need to build that expertise.

Tom Temin: And when you mentioned the case of this employer that kept having this bias toward white men in hiring, because that’s what they had hired before, that gets to the issue of the data used to train algorithms. It seems like a big clue to a lot of these questions about AI, in whatever domain, is the data that you train your algorithms with. And it seems like that’s a good place to focus?

Charlotte Burrows: That’s right. And that is really a key part of this. I would say that there are also other issues that are particularly important for employers to be thinking about. One of them is that we have, if you well, sort of gamified assessments out there. So it’s basically you take a job test, that is something like, you know, almost like playing a video game. And in some instances, those may require a certain level of dexterity if there’s like, say, a keyboard or a mouse or some other sort of manual input device. And in those instances, it’s really important for employers who want to use those kinds of tests to be thinking about whether or not they actually need to offer some accommodation. So if you had an applicant, for instance, that really needed assistance, because of limited dexterity, they might be experts in whatever it is the job was, but just had a little trouble or needed extra time to use whatever manual input device, the trackpad, etc. It really is important that employers make those accessibility decisions and offer those accommodations so that applicants have the opportunity to assert that that’s necessary. And those are things that are fairly easy for the applicant to know about. There may be other things that are embedded that the applicant doesn’t know are part of the assessment. And so that’s why I think it’s important for us to be particularly vigilant as the federal government in helping as well,

        Sign up for our daily newsletter so you never miss a beat on all things federal

Tom Temin: Yes, and speaking of embedded information that can work for or against a person, I heard a case recently of an agency that was trying to make sure there was no bias in its grant program. And it learned that it doesn’t ask racial information on grant applications. I said, so how can you tell? And the answer was, well, we did a name search across the applicant pool. And that seems to imply that certain names are associated with certain types of people. Is that reality or bias? Or what happens in a case like that?

Charlotte Burrows: In certain instances, that’s true. There are certain names that grants will be, for instance, for immigrants or for persons from particular ethnic groups. Obviously, it’s not, you know, sort of Spanish surnames, sometimes you absolutely can tell, but not with complete accuracy under any circumstances. So without knowing more, I don’t know that I can comment in any detail about that. But what is important to understand is that these kinds of searches, if you will, inadvertently can end up excluding, for instance, if you had, again, trained your AI tool on, you know, past applicants who did not have anyone with a Latino or Spanish language surname, then you might actually inadvertently end up excluding people who had such surnames in it. So there are certain names, it’s certainly not a perfect match — certain names that are associated, rather, but not a perfect match.

Tom Temin: Is there a timeline on it? Does it have an end date? Do you have deliverables that you’re planning on, or is this going to be something you think might be just a continuous process?

Charlotte Burrows: Well, it’s definitely going to be a multi-year effort. This is a very important and, if you will, sophisticated area of the law. So we’re looking at doing that thoughtfully. And a couple of our goals, really, as we think about it is to really increase transparency around this. And to encourage employers to be really looking, if you will, under the hood, when they seek to purchase these products. Most employers don’t develop them themselves, and to make sure that they’re actually valid for the purposes that they are looking into them for. So we’re going to be launching this series of listening sessions with key stakeholders, and looking towards gathering information and identifying promising practices. And I think that’s really where we can be most helpful as this area develops. And to sort of give some bright line guidelines based on the conversations and the research that we will do, and our expertise in the civil rights laws, to really sort of assist in that area as the technology develops. I think it’s sort of the sweet spot of where we can contribute.

Tom Temin: And while we have you, you have been recently just a few months ago I guess, well, beginning of the year now — it seems like a million years ago — reconfirmed now and you’ll be at the EEOC for a few more years. Any other big initiatives you want to talk about while you’ve got the chance?

Charlotte Burrows: Oh, thank you for that question. So we have a lot of issues on our plate always, as I’m sure you’re not surprised to hear. And a couple of things that have been different about this past year and certainly the year since I became — or almost a year, I guess, not quite — since I became chair is that we are looking to really assist in the area and respond to this big overarching issue of what has happened in the aftermath and the ongoing circumstances of this pandemic. So it turns out that the COVID-19 pandemic has been not just a public health crisis, and a crisis with respect to some economic issues, but it’s truly been a civil rights crisis, because it has hit so heavily on people of color, on women, certainly women with children, persons with disabilities, to some extent the LGBTQI population. And so what we need to do is really be laser focused on ensuring that as we come out of, you know, this pandemic economy, we are including everyone. So that has been an overarching thread. We had a tremendous hearing back in April to really focus on those issues in a broad way, and are continuing to do that, as well as to push out really detailed information about how to deal with the civil rights implications of the pandemic, including things like accommodations for persons with disabilities, and accommodations for persons who have religious objections, in some instances, to save COVID-19 vaccines or the kinds of things that just that employers need to be aware of, as they wrestle with what’s really been a pretty big challenge for employers throughout that period of pandemic, and will continue to be, I think, as particularly as some of those remote workers return,

Tom Temin: Just going to say that you make a good point there. I spoke to a gentleman just the other evening who is African American and employed and has had the COVID in a pretty severe way. And coming out of it, he is fine. Now he’s totally mentally there, always was. But there are some lingering physical issues that could almost render him not disabled, but needing accommodation. That’s the kind of thing, I guess, that you’re looking at to make sure employers are open to accommodating and, in fact, accommodate?

Charlotte Burrows: That’s right. And so there are a host of issues, but this is also one that we want to be really educating about because COVID-19 itself, or the effects of it, can even if it’s not that way in the beginning in certain circumstances, and I wouldn’t be careful about that, because it’s not always the case. But in certain circumstances, COVID-19 itself can be a disability. And that’s something that we’re looking to providing additional information about in the near future. Our partners at the Department of Justice and Health and Human Services have also spoken to that issue. And we really want to help, with respect to the employment context, employers to understand this and to, you know, support employers like the one that you just mentioned, as they recover and so that they can fully work and really contribute all of their talents in this area. The other thing I would say is that retaliation, including unfortunately in the federal government, has proven to be a huge issue. And so we joined with the NLRB and with the Department of Labor recently on a broad initiative to educate first of all the employer community but also civil rights groups and employees about how to tackle that. We take retaliation so seriously, because it’s really, if you think about it, the linchpin that runs through all of our other initiatives. And if you don’t have the ability to, you know, report discrimination without fearing retaliation, then that right to be free of discrimination really only exists on paper. So that’s a top issue for us as well.

Tom Temin: Sounds like a full plate.

Charlotte Burrows: Absolutely, always. And we are really enjoying it. So delighted to have this opportunity to talk about our work.

Tom Temin: Charlotte Burrows is chair of the Equal Employment Opportunity Commission. Thanks so much for joining me.

Charlotte Burrows: Thank you.

Copyright © 2024 Federal News Network. All rights reserved. This website is not intended for users located within the European Economic Area.