To anyone who doesn’t dream of labor, the idea of a personality hire — a person who gets jobs because they’re a fun, unique human to be around — can be a bit jarring. And there’s an idea that might be even more puzzling than the existence of personality hires: that these people can teach us valuable lessons about being hired today. It would behoove us to start thinking about the job process through the eyes of a charming, pleasant, fun, distinctly human personality hire.
Statistically, today’s job market is not great. But more than that, the inevitable creep of AI has made the process of applying for jobs even more of a headache than it was before. Never has the application felt more crapshooty (and crappy), when it seems like so many recruiters and employers are using AI to sift through applicants. The idea that someone’s life-changing opportunity could all depend on the way a computer responds to a prompt feels equal parts dehumanizing and maddening.
Even more demoralizing is when you realize how fickle and faulty AI can be.
A study from researchers at Columbia Business School tested three generative AI models — GPT-3, GPT-4, and Llama 3.1 — and found that, when there are multiple options (i.e., a person in HR plugging in multiple candidates), the AI models can show “order bias” and will default to the first option (candidate) listed. Olivier Toubia, a professor at Columbia Business School and one of the authors of the study, talked to me about the worrying results and also shed light on how he thinks we can be more successful in the wake of an AI-powered job market.
Essentially, the more the job market turns to AI, the more we should think about how to be more human to stand out. And, at least to my mind, no one may do that better than personality hires. This interview has been lightly edited for length and clarity.
Olivier, the impression I’m getting is that there are parts of the job market where both employers or recruiters as well as applicants are using AI. Recruiters are using AI to screen resumes, while candidates are feeding their resumes and CVs to the machines to better their chances.
Yes, we’re starting to see applications of AI for hiring and for screening candidates. We’re starting to see AI agents processing CVs, and actually I think that’s a bit concerning.
I’d assume that jobs are important and people don’t necessarily want a machine making that very crucial decision versus a human. Tell me about your research.
What we found in our research is that [if you’re using AI], the way you order the options may influence the results. If you say, Here’s A, B, and C, which one do you think is more qualified? It turns out that the way you ask the question and the way you label and order the options [the candidates being recruited] is actually going to influence the results.
We found systematically that which one is labeled A versus B versus C, and how we order them impacts the choice that GPT makes.
So people could be picked for an interview or a job just because they were entered as the first option, and AI may be more likely to go for the first option just based on numerical order.
Yes, and I think it could be a bit dangerous because you could actually end up with the results that are going to be driven by pure random factors. What was the order of the candidate on the list as opposed to the actual qualities of the candidates? That’s one issue.
We also found that there’s potential stereotypes or biases that might come up. Maybe some candidates might be judged as being more or less worthy based on characteristics that should not be considered, [like] identity, gender, ethnicity, age, or other factors. There’s the risk of possible biases, social biases, and just also idiosyncratic biases. That’s just one issue of using AI as a recruiting tool.
How concerning is that for you? I don’t hire people. I don’t fire people. That’s not my job. That said, to me, I think it sounds alarming that AI could be basing such a crucial decision about my future based on what order I was entered into a system.
Generative AI is impressive. It’s able to solve complex mathematical problems. It’s able to write poetry. It’s able to do things that only very smart humans will be able to do. So because of that, we tend to think that gen AI is superhuman. But it’s not like a human brain. It’s a very different architecture.
We want to assume that we can trust Gen AI, and that we don’t need to worry about simple errors that it could make. So that’s why maybe we tend to turn a blind eye on this. And I’m actually surprised that not more people are aware of these biases. It’s an inconvenient truth.
I want to ask you about the idea that AI “levels the playing field” at work.
That’s been a debate, in academia and in practice, of augmentation versus automation. Is AI going to make us more productive? Is AI going to automate us? Is it going to level the playing field or is it going to actually amplify differences? I think so far we don’t have a lot of evidence that really leads to very clear conclusions.
My own take on this is that in many cases, it might level the playing field and indeed helps people who maybe have less skills and less experience perform similarly to those who have more. It can level the playing field among humans. But then the issue is: could an AI then do the job without humans at all?
Right. If you’re totally reliant on AI, then what exactly are you doing?
Generative AI may level the playing field, but AI is also going to replace humans because if AI can do well enough, then why would anyone hire a human and pay a human to do something that AI can do well enough?
Yeah, I know. In my opinion, the people who are going to “win” with AI are the ones who are going to use AI as a source of differentiation to differentiate themselves. And there’s two ways to differentiate yourself. There’s vertical and horizontal.
Vertical means being better. So basically: Those [who] are going to use AI to be better than they are without AI, but also better than other people with AI, and also even better than AI without humans. If you have the skills and the experience and use AI to be the best performing agent, then I think that’s going to become actually more desirable on the job market.
And what does “horizontal” mean?
It means being different. It’s more about having a unique personality, unique points of view, and then maybe using AI to really express your ideas and your views in more compelling ways.
If the job market is people using machines to look for candidates and people using the same machines to look for new jobs, then it sort of becomes two machines talking to each other, right? One of the few differences between candidates may be having personality, right?
I see some use cases where maybe people use AI to create more compelling work products that actually are going to really showcase their unique ideas and personalities and points of view. And that I think is another area in which maybe AI could actually help humans. If you’re using AI, you want to use AI to differentiate yourself either by being better or by being different.
That’s interesting: Being different could be as important as being better?
It’s two different pathways in some cases. It’s not like one is above the other. It’s more like there are two ways that are not exclusive in which people can differentiate themselves.
Do you have an example of a field that values horizontal strength?
Any field that has to do with creativity, opinions, and personal connections. Any field where it helps to really have a connection with your audience. Maybe some consulting industries and creative industries where people will pay for unique insights, unique perspectives, unique points of view that are very sharp and distinct from others.
People have always told me that when you write cover letters, you need to make sure to stand out and tell whoever is reading why they need to choose you. That seems like it’s more important than ever in this job market. We have to really let ourselves shine through, otherwise we risk getting lost in the shuffle.
Yeah. I told my students, “Don’t be afraid to make yourself heard. Stand out, even be annoying — be a human.” You want to actually stand out.
Okay, maybe not when you write your cover letter, but in life, you want to stand out. You want people to notice you; you want to have something to say. If you just blend in, then you will just be replaced by AI.
By the way, there’s now also AI tools to help you with live interviews when you’re on Zoom. There’s an AI that is actually helping you answer questions live. So you basically have to have the person live in front of you in your office with no electronics to be able to really test them
The whole idea just makes me sad? But it also kind of makes me a little hopeful that maybe the only way to succeed is to be a person.