It’s fair to say that there has been a somewhat mixed reaction to the revelation earlier this week that language learning app Duolingo has been using a ‘taxi test’ to vet potential new staff. Speaking to The Burnouts podcast, Duolingo’s boss Luis von Ahn, who employs more than 800 people, said, “Whenever you’re trying to hire somebody it’s because you have a hole in the organisation you are trying to fill. The informal rule is ‘better to have a hole than an asshole’.
“You can’t always tell from the interview. Part of the interview for us is how they treated their driver. Our belief is if they are gonna be mean to their [taxi] driver they are probably gonna be to other people, particularly under them, so we did not want that.”
Duolingo’s approach has understandably raised questions about how firms select staff. A number of tech giants (not least Google, which once used cryptic, anonymous billboards in Silicon Valley to recruit engineers) are known for their rigorous weeding processes, often involving complex tasks and problem solving. Others firms are increasingly farming out the hiring process (or at least its early stages) to artificial intelligence. Others still call in potential employees for interviews at weekends, to gauge how willing they are to work out of hours. By this measure, Duolingo’s ‘taxi test’ is mild.
Beyond the novelty of the ‘taxi test’ there’s also a question of whether or not recruitment processes that include such tests are anything other than performative? Do firms aim to gain exposure from such practices, or are there real benefits from hiring only staff who are nice to their taxi drivers?
Von Ahn clearly thinks so. He told The Burnouts that Duolingo had once spent the best part of a year hunting for a chief financial officer. The hiring committee loved the final candidate. Strong CV, aced the interviews. But the driver reported back that the candidate had been rude during the ride from the airport. Duolingo walked away from the hire.
In the kitchen at parties
He is not the only boss with a taste for covert character assessment. Trent Innes, formerly managing director of accounting platform Xero and now chief growth officer at SiteMinder, has his own version: the coffee cup test. Innes walks every candidate to the office kitchen for a drink before the interview. At the end, he watches to see whether they offer to take the empty cup back. Those who leave it sitting on the desk don’t get the job. “You can develop skills, you can gain knowledge and experience, but it really does come down to attitude,” Innes told The Ventures podcast. Not everybody agrees. Social media users accused him of playing “deeply disturbing psycho-social” mind games. One pointed out that most guests in an unfamiliar office would not presume to wander into the kitchen uninvited.
The backlash highlights a tension at the heart of these theatrical hiring tricks. They make for terrific podcast anecdotes. They generate column inches (this article being something of a case in point). They project a carefully curated image of a company culture that values decency over credentials, which is all useful PR for firms competing for talent in a tight labour market.
The trouble is, nobody has produced a scrap of evidence that a candidate’s behaviour in a taxi (or towards a coffee cup) predicts anything meaningful about their performance on the job. A nervous introvert might sit in silence during a cab ride and go on to be a brilliant engineer. A polished sociopath can be perfectly charming to a driver and an absolute nightmare to manage. Von Ahn’s maxim about holes and certain orifices is catchy, but it’s not science.
There is, at least, a strong financial case for being picky. The US Department of Labor reckons a bad hire can cost up to 30 per cent of the employee’s first-year salary, and some three-quarters of employers admit to having made a wrong hire at some point. Getting it right matters.
A job for AI?
Which is partly why so many firms have turned to algorithms for help. Some 87 per cent of companies now use AI somewhere in their recruitment process. The technology screens CVs, ranks candidates and even conducts preliminary interviews (AI-led interviews have more than trebled in two years, to 34 per cent of initial screenings). Proponents say it cuts time-to-hire by a third.
The track record, though, is patchy. Amazon famously scrapped an automated recruiting tool in 2018 after discovering it had taught itself to penalise women. The system had been trained on a decade of CVs, most of which came from men, and duly learned to downgrade any CV containing the word ‘women’s’ or the name of a women’s college. Two-thirds of job-seekers say they would not apply to a company that uses AI to make hiring decisions (although how they might know, or find out, is a moot point). The EU AI Act now classifies recruitment AI as high-risk, with compliance obligations kicking in through this year and next.
All of which brings us back to the taxi. Perhaps the appeal of Von Ahn’s test has less to do with its predictive power and more to do with its stubborn humanity. In a hiring landscape dominated by algorithmic screening and AI-scored video interviews, the idea that a CEO still cares how you treat the driver feels almost quaint. The coffee cup, the cryptic billboard, the weekend interview: these are signals that a human being is paying attention, not just a language model scanning your LinkedIn profile.
Whether that attention translates into better hires is anybody’s guess. But as a recruitment strategy, it certainly beats asking a chatbot.
Photo: Dreamstime.
