Recruiters Missing Signs That Tech Job Applicants Are Using

BY KIT EATON@KITEATON

Tech workers are facing mass layoffs, at least in part because the exploding AI revolution is rapidly shifting some work from humans to artificially intelligent software that performs sometimes slow and grinding tasks automatically, in less than a second.

Now, in a decidedly circular cause-and-effect loop, some of those laid off workers seeking new employment are using AI software as part of the application process when they're seeking new jobs--and the human resources personnel guiding the hiring process can't really tell the difference. 

Business Insider reports that this is particularly true regarding coding, an important part of interviews for some highly technical jobs. Application questions like "write code in Python that turns 'hello world' into a 3D playable game" or a similar tricky task are designed to put a prospective coder employee under pressure and test their skills. Exactly like "explain how you'd solve this thorny office situation" questions might stress an employee destined for a less technical role.

An experiment by industry recruiting platform interviewing.io, cited by Business Insider, shows it's extremely easy to just ask an AI chatbot like ChatGPT to solve this sort of coding task for an applicant. Worse, no interviewers in the experiment could tell when a candidate had cheated.

That survey was based on just 37 job interviewers, though. The small sample size means it may not meaningfully scale out to apply to other high-tech job interviews that require coding. The experiment also found that when the interview problems were written to be more unique to the needs of a particular job or industry, ChatGPT-based code was less likely to give the right answer. Code tasks taken from an online service that sets coding problems were easily solved by ChatGPT about three-quarters of the time, but just 25 percent of ChatGPT answers were correct when applied to more cleverly hand-crafted code problems in interviews. 

This part of the interviewing process leaves hiring managers with a similar challenge to devising a really specific interview problem in a more traditional question-and-answer situation. The problem, or question, must skirt pre-researched answers that an interviewee may have already prepared to answer. Since this part of the hiring process requires an employer to try to work out if an employee really has the skills and savvy to successfully tackle the job, it requires some work on the company's end as well.

This is where the notion of using ChatGPT to "cheat" on a high-tech interview begins to show logical holes. Developers of all sorts now regularly rely on other people's code, and AI-crafted code, to complete their projects. It's a natural part of today's startup software-writing world, where vast repositories of coding knowledge are available online in the form of systems like GitHub and chatbots are freely accessible. Knowing how and where to look for help with a tricky coding problem, be it quizzing another expert on GitHub or asking a chatbot exactly the right query to get it to produce AI-written code snippets you can easily integrate into your own code, is a necessary skill.

AI-assisted cheating has also been a recent topic of many a hand-wringing headline like "56% of College Students Have Used AI on Assignments or Exams," as AI takes a role in "helping" students complete their studies. And though some late 2023 research from Stanford University, reported in the New York Times, showed fears of high school students using ChatGPT to cheat on assignments were overblown, these worries are only likely to remain: Everyday AI use is only growing, and Microsoft is even building AI assistants into the very productivity software we all use to create documents for work--and essays for school.

The lesson for high-tech recruiters here is that they need to modify their interviewing habits for this new world. It's even possible an AI could help with that.

Photo Credit: Getty Images.