How AI is Deciding Who Gets Hired

The great resignation is continuing to loom. There are more than 11 million unfilled jobs right now. We hear from businesses all over the country, that it's very hard to hire people. Almost two years into the pandemic, workforce participation remains slumped, while job openings are near record highs. And, this is coming at a time when the recruiting process is going through a radical transformation. Companies in the past who were maybe getting dozens of applicants for a single vacancy. Now, they

're getting hundreds, maybe even thousands of applications. And, so, software has come in to automate the process of looking through all of these different candidates. If you've applied for a job in the last five years, you've probably been impacted by this technology, whether it was screening your resumes, making you play games to analyze your personality, making you go through some other automated system. And, some people are fighting back. Who's accountable when those AI machines break the la

w? Why it's important that Congress take action to ensure that big tech is held accountable. They claim there is labor shortages, but there is a plethora of people out there who are looking for jobs . We're at a moment of growing crisis with AI hiring. And, that's why people are eager to act. Amidst this whirlwind of disruption, one has to ask the question, how did we go from this to this, to this? More than a decade ago, instead of just replying to a job in a newspaper, printing out hundreds of

resumes, and sending them out physically to companies, now you can apply to even more companies all at once by doing it over the internet. If you hate going to work and your coworkers don't respect you. Hey, dummy. It's probably time - you need a new job. Companies now, they're getting hundreds, maybe even thousands of applications. And, so, software has come in to automate the process of looking through all of these different candidates, effectively using something called ATS or applicant trac

king systems. These systems effectively track different applicants, look through their resumes, and take them even potentially through the whole process. They will take candidates who are at the top, the middle, and the bottom of performers as deemed by a particular company or hiring company, and then train their systems based on those judgments. And, then, use that as a baseline on which to make it's own automated judgements about whether someone is good or bad. Not just students, experienced p

rofessionals. They just don't know that these tools are used so frequently. Gracy Sarkissian runs New York University's career center, where she gives tips to optimize applications for AI systems. Students will encounter AI tools in an interview, if they're doing a virtual interview. Some folks will also counter AI powered assessments as part of their interview process, depending on the industries in which they're entering. Well, the use of the software is very prevalent. There was a recent Harv

ard business study that came out showing nearly every single one of the Fortune 500 companies use this kind of software. Large companies are never gonna be able to stop using this. They just have no choice, because of the inbound flood of applicants they get every single day. Most vendors position their software as something that can alleviate the human bias, which has shown to be endemic to the hiring process. But, critics question whether AI is actually injecting fairness or just making those

same biases more efficient. I definitely think there is a segment of the population that is being excluded or not fully represented. And, they are being left to the wayside. This is Jamaal Eggleston. He's a work readiness instructor at the Hope Program, a nonprofit that helps New Yorkers find jobs. My students aren't prepared for it. And, coming from a population that might not be, again, as familiar with the technology, they run into pitfalls, and they don't do as well as other communities when

encountering these AI. Since US companies don't have to disclose when they're using AI, many applicants are still writing resumes catered to the human eye, rather than to AI. When people tell you that you should dress up your accomplishments or should use non-standard resume templates to make your resume stand out, when it's in a pile of resumes, that's awful advice. The only job your resume has is to be comprehensible to the software or robot that is reading it, because that software or robot

is gonna decide whether or not a human ever gets their eyes on it. Ian Siegel runs ZipRecruiter, one of the dozens of AI companies which transformed the job hunt in recent years. You want to write your resume for robots, not for humans. These softwares not only review them to decide who gets passed to a human, but another thing they do is they write simple summaries of who you are for those humans, so that those humans may, in fact, never see your actual resume. And, it's not just resume screeni

ng that has changed. Initial job interviews and assessments are being automated at companies like Goldman Sachs, Delta Airlines, and Walmart. This screen may look familiar. You have a minute or two to answer questions or play an assessment game, which will determine whether or not you move on to the next round. So, we decided to try out one of these systems for ourselves. Talk about a time when I was overwhelmed or stressed. I had an employee who's working style and my management style were not

in sync. Alright, playing the first game, it's kind of like a remembering, memorizing. Four, six, seven. And, you have to, okay, so it's like pattern recognition. Oh. Okay, maybe I'm not getting the job. I don't know how much this is going to tell about a person. What? Okay, your assessment has been submitted. Please, click the button above to continue. While our tests were unscored, many companies algorithmically rank interviews based on word choice, vocal tone, and facial expression. A lot of

people are missing out. A lot of people are not giving back what the interviewer is looking for, because it was programmed a certain way. These things are programmed by a certain segment of the population that might not be totally inclusive or not be fully aware of how other cultures emote. I believe race is a part of it. I know it's trying to be... Trying to leave race out of it. I know it's trying to be not biased, but some things are inherent, even if it's unconscious. I like to call it the g

host in the machine. In a statement, HireVue said its systems are rigorously audited for adverse impact, meet all regulations, and only analyze word choice. Last year, the company announced it discontinued facial analysis, in part over bias concerns. Applicants trying to navigate this new world of hiring find themselves looking for help on social media. This is how to beat automated resume screening software. Here's how to trick the online job application. If you know how to play the game, you'r

e actually able to stand out on top as the best candidate. One controversial strategy is copying an entire job description into your resume, hidden with white font, in order to trick the AI into picking up keywords. They won't be able to see that you have all of those qualifications and its exactly what they have on their website, because it's all white. It's eventually going to go to a human being, that's going to look at that resume, and say, this is ridiculous. It's not about beating the syst

em. I think that that thinking is not gonna help candidates to be most successful. What helps you be most successful in the job search is to know that the world of work is always changing. There are tech vendors that help candidates optimize their resume for systems. We have a tool that we use at NYU, other universities use it, it's called VMock. Students can put their resume into the system and it will give them a score. And, one of the things that it's scoring for is how likely it will be to b

e picked up by an ATS for a particular industry. While elite universities have access to resume optimizing tech, others may be left behind in this new world of work. There are millions of people who are being left out of the job market, because hiring software is discounting them very early on on the process, when it scans their resumes. We have students who... Some had bouts with incarceration, some of our students are college students, some of our students are long-term workers who have been l

aid off because of the pandemic. And, these are good people who have great experiences, who can add and become assets to a lot of these companies. For example, a person who, perhaps, spent some time in the military or someone who went on parental leave might have a year or two long gap in their resume. Now, if you were going to interview with a human being, you might be able to explain why that gap is there, but an algorithm doesn't necessarily see that. It just sees that, between a couple of ye

ars, you weren't working. A lot of times, they get no response. And, it's really hard, because there are... They don't know what they did wrong. They don't know if they did anything wrong. There's this term, the black box of artificial intelligence. What's happening with some of these algorithms is they might be discounting hundreds of different types of people, but even the people who create the software don't know why. In a statement, icims, one of the leading ATS vendors, said it takes these

concerns seriously, and has a team focused on responsible AI, and follows industry initiatives on fairness. Algorithmic discrimination is difficult to prove, but it's been revealed by audits. Amazon scrapped its AI recruitment tool in 2018, after it discriminated against applications which listed women's colleges or sports. In another system, the best predictors of job performance included being named Jared, or having played high school lacrosse, according to an auditing law firm. The algorithms

that are being used to look at people's resumes, to conduct interviews, they've been trained on data about previous applicants, is over perhaps the past decade. That's how machine learning works. No one explicitly told these systems to discriminate, but they were trained on millions of biased hiring decisions. The fear with these algorithms is that you take bias data in, because they're using the wisdom of the crowd and the crowd is clearly biased, and what you're really doing is perfecting bia

s. That's the accusation that's been made. The reality is that when we deploy these algorithms, we, as the software engineers and programmers, get to make choices about what data the algorithm can and cannot use. We can say to the algorithm, you never get to see someone's name. If I tell you someone's name is Jeremy Levine. You know something about them. If I tell you somebody's name is Natasha, you know something about them. And, by removing them, we create the most level playing field we possi

bly can, so as much as possible, it's a merit-based system. While automated hiring tools have become ubiquitous, public perception hasn't caught on. Recent Pew surveys found that most Americans didn't know resumes are being screened by computers and most considered the practice unacceptable. But, one crucial group is paying attention to AI, lawmakers. The European union has unveiled its proposed strategy for artificial intelligence, including regulations on transparency and data protection. Conc

erns bout how this technology could exacerbate discrimination, for the FTC, I think, foremost, the FTC needs to be making sure that we're fully understanding this technology. A federal civil rights agency is scrutinizing automated hiring tools, while Illinois and DC have explored some of the nation's first AI hiring regulations. The Federal Trade Commission recently gave a message to companies using AI, hold yourself accountable, or be ready for the FTC to do it for you. In New York, we just ena

cted one of the first AI hiring regulations in the country. Albert Fox Cahn runs a technology advocacy group and participated in New York's task force to understand the impact of AI systems. The bill basically started off by saying, before you sell this technology, you'll have to do an audit, and you'll have to tell people when this technology has been used. And, when you look at what this bill did, it was regulation in name only. Having an audit can be useful, but when that audit is done by the

software vendor themselves, well, that's not worth the paper it's written on, and it's absurd. But, in the New York law, you can't sue. It doesn't matter how egregious the software is. It doesn't matter how biased. You don't get your day in court. And, without that, none of these bills can work. In response, some of the world's largest companies like IBM and Facebook parent Meta, have devised an auditing standard for AI and hiring. However, critics say that industry standards can't replace true

government regulation. We don't trust companies to self-regulate when it comes to pollution, we don't trust them to self-regulate when it comes to workplace comp, why on earth would we trust them to self-regulate AI? Look, I think a lot of the AI hiring tech on the market is illegal. I think a lot of it is biased. I think a lot of it violates existing laws. The problem is you just can't prove it, not with the existing laws we have in the United States. So, a big part of this is playing catch up

to open up discovery, to open up the court process, to say you can look at these algorithms more easily. With widespread adoption, the use of AI hiring tools is less the future of work and more like the present. Virtual assessments, interviews, and job applications likely aren't going anywhere, but it's up to the employers to decide how much of their process they really want to automate. Smaller companies have a little bit more choice here, because they don't get as many applicants as the big c

ompanies do. So, one thing that's been suggested to me by people in the recruitment industry is actually maybe the small to medium-sized businesses should just be a little bit more critical in their thinking of whether they really want to use this software. Because, in some cases, just using a human being to interview someone, even to look through all the resumes might stand a better chance of finding that star, that star employee that they're looking for. Keep a segment or a avenue open for an

older person, for a person who was incarcerated, for a person who's young, for a person who's just not aware of these technologies, so they can also have an opportunity to get hired. We can't solely base something as important as finding a job, as finding a career, to only to AI. It's hard to fight against something that has been programmed one way or the other. We can only teach our students how to navigate these waters.