why a.i. probably won’t find you the perfect new job
Interviews are the most important and uncomfortable part of the hiring process on both sides of the table. As a job seeker, you’re trying to figure out what will impress your potential future bosses while looking for red flags. As an employer, you’re trying to decide if you can gamble on the person in front of you and figure out whether any gut feelings you might have are in any way legitimate warnings. Add shrinking training budgets, if they even exist, and stagnant wages, and you can see why record numbers of jobs are now going unfilled while a record number of workers are underemployed and pundits are ready to throw in the towel on the notion that jobs being lost to automation will simply be replaced with new careers.
The solution, according to some futurists and entrepreneurs, may be to let artificial intelligence match employers and workers by collecting as much relevant data as they can and trying to get as close to a perfect match as possible. And the CEO of Woo, a company which is trying to make this arrangement a reality, is optimistic that the process will tell employers so much about their potential hires that interviews might become things of the past. All it will take to find a job is to take some relevant automated tests, submit a resume, and receive offers in your email, while hiring will be as easy as posting an opening and receiving a ranked list of potential workers. Or, as he told Forbes…
If you think about an interview it’s an outcome of a lack of information on both sides. They [candidate and employer] have to talk with each other in order to understand what you know and what you don’t know. But if there’s a machine that knows everything – like a god – knows about your past experiences, about your projects, your culture – the machine is going to tell you that there’s a perfect fit.
But the problem is that we don’t have a machine that knows everything and what Woo offers is just smarter keyword matching code that allows for some variation in people’s resumes. It’s not going to be able to identify what constitutes a transferable skill because even employers today aren’t sure or seem to care very much about them. All it can do is crunch through the kind of data about an existing workforce at a company and identify more candidates that seem to be in the same mold, promoting the exact kind of corporate monoculture that leads to groupthink and lackluster results which already has companies complaining that they can’t find the right people while simultaneously being unable to clearly articulate what it is they want.
Even worse, unlike the web’s regressive pseudo-intellectuals will tell you, bad data can lead an artificial intelligence to turn downright bigoted and discriminatory, something Amazon found out the hard way after having to scrap an AI that outright refused to recommend any women who applied for jobs with the company. Instead of learning what skills matter to Amazon, the AI went after the gender imbalance it saw and assumed that indicators of gender rather than skills themselves, were the most important part of the screening process. And this is not surprising if you know how artificial neural networks work. If you give it vast reams of data, it will latch on to the most consistent trends and prioritize them as critical.
And this really gets us to the heart of the problem. When companies rely on AI to absorb the resumes of existing employees and come up with criteria for hiring new ones based on their existing workforce, they’re making the implicit assumption that those working for them now are the best possible people for the job, even if that’s not necessarily true. Any hidden bias in their hiring practices is going to be learned and magnified by the neural networks they’ll employ, and as a result, they’re not going to end up with new workers as much as new versions of people who already work for them because that’s who the AI will consider the perfect candidate. So, in a way, having a computer attempt to pick your new employees is a good way to figure out where you are overlooking potential new talents, but a bad way to do actual, honest-to-goodness hiring.
Every unconscious bias an employer could have is just going to be exacerbated by technology designed to learn how to repeat past decisions in the span of milliseconds. Rather than opening new horizons for workers and introducing employers to new ideas, these AIs run a very serious risk of further pigeonholing millions. This is why hiring for creative, collaborative, open-ended tasks which will have to be done by humans even in a future where every possible bit of routine and drudgery has been automated, is a human problem, not a mathematical one. And instead of looking to cut corners, employers need to figure out their vision for the future and actively seek out new and creative pools of potential workers to help them make it a reality.