You can almost hear the howls of frustration from HR departments. Jobseekers have discovered artificial intelligence and they’re not afraid to use it. Employers have become snowed under by people using the new tools to churn out impersonal applications. Some applicants are using AI to bluff their way through online assessments, too. The FT has reported that many large employers have a “zero-tolerance attitude towards the use of AI”. 

I’m sure that would be news to job applicants, who have had to put up with the use of AI by large employers for years. Indeed, jobseekers would be well within their rights to say: but you started it.

Like many cautionary tales, this one begins with good intentions. In the 2010s, employers implemented new automated recruitment tools to whittle down candidates before the interviews because they wanted to make the process more efficient and more fair, with less risk of human bias.

“Asynchronous video interviews”, for example, involve job applicants answering questions alone in front of their webcams with no human on the other side. Often, an AI system assesses their responses. But I have never met a jobseeker who liked them.

In 2021, I wrote about research which warned that young people felt confused, dehumanised and exhausted by the new tools. I was flooded with responses. “I did one of these and it was the hardest and most humiliating experience I’ve ever encountered,” one older man wrote. “An interview itself is hard enough for someone who is in the job market for the first time in years, but then you throw this at them. I don’t mind saying I was traveling and did one from a hotel (not an ideal set up with my iPad balancing on a suitcase) for a company that I had dreamed about for 30 years. The pressure was so intense, I messed it up.” Afterwards, he said, he sat in his hotel room and cried.

So it should be no surprise that jobseekers have turned to new generative AI tools such as ChatGPT to speed up or “game” a process that already felt dehumanised. Videos have even appeared on TikTok in which people demonstrate how to use ChatGPT to provide answers to questions in asynchronous video interviews, which the applicant then simply reads out. 

That said, the ensuing AI arms race is clearly not going well for anyone. It was supposed to improve efficiency and fairness. It has become a threat to both. 

On the efficiency side, employers complain they are overwhelmed by applications, which only fuels more rejections. “I hesitate to say it breaks the system, because it’s not broken, but it does mean you’re getting ever more applications coming,” Stephen Isherwood, joint chief executive at the Institute of Student Employers, told me. It has also become harder for employers to find the best candidates because some applicants are using AI to boost their scores in tests of their skills.

Uneven access to the paid-for AI models that perform best on hiring assessments might be fuelling a new type of unfairness, too. Jamie Betts, founder of assessment company Neurosight, told me their survey of 1,500 jobseekers last year found that 31 per cent of men were using a paid-for AI tool, compared with 18 per cent of women.

It is worth noting that Neurosight sells a tool it markets as ChatGPT-proof, so Betts has some interest in making this point. Nonetheless, he said his company had recently completed an analysis of a well-known critical reasoning test for a global professional services firm. “Year on year, what we’ve seen is significant increases in [relative] underperformance among black individuals, females and [those with] socio-economic indicators like whether you had free school meals.”

What are the solutions? Some online assessments are, for now, less vulnerable to AI use, such as those that involve playing short games. But I wouldn’t be surprised to see the return of mass in-person test centres for technical skills assessments. Isherwood and Betts both said employers were thinking about reintroducing the human touch earlier in the process, too.

Even HireVue, a big vendor of asynchronous video interviews, wrote in a paper last year: “One of the best ways to reduce cheating behavior of all types is to utilize a multi-stage workflow in which a candidate’s knowledge, skills, and abilities are validated by an interviewer in a live interview setting.” 

Would there be a trade-off with bias and consistency? Perhaps. But if there is a lesson from this cautionary tale, it is that technology can’t just magic trade-offs away. Seemingly simple solutions to hard problems don’t usually stay solved for long.