This post was originally published on this site.
At the weekend a friend showed me a message they received after applying for a job.”We want to inform you that we have completed the review process of all received applications, and have moved forward with other candidates who more closely meet the specific requirements of the position at this time,” the email read.
This message was sent to the job seeker a few minutes after submitting their job application for a local company where job applications closed in October and it begs the question – how did the hiring firm review the application so quickly?
That answer is quite simply artificial intelligence.
The current most fashionable tech on the scene is drawing in billions in investment as Silicon Valley desperately tries to get something “new” to stick. Cryptocurrency and the blockchain failed, non-fungible tokens failed and the metaverse failed. Now, big tech has loaded its eggs into the AI basket in hopes that business will bite.
While AI has many uses, the core of the technology’s promise is that it will make repetitive, mundane tasks easier to complete. This means off-loading tasks that a human once did to a machine and problems inevitably pop up because, frankly, the technology is problematic and prone to mistakes.
However, this hasn’t stopped companies from implementing the technology and concerningly, it has found its way to the hiring sector.
We’ll be honest, combing through CVs can be a daunting task, Vervoe, a company creating AI-powered hiring solutions, reckons companies can spend 23 hours scanning CVs to fill just one position.
“The rise of high-volume hiring has made manually managing the resume screening process virtually impossible. And, as more recruiters start to realize that unconscious bias is impacting their hiring success, companies are turning to machine learning and recruitment software to help with screening candidates fairly,” writes Vervoe.
Except it may not be doing that. Social media is awash with stories of job hunters stumbling at the first block because their CV was seemingly reviewed by an AI bot.
The problems with AI
AI is only as good as the data it’s trained on and that presents a problem for South Africa.
Most, if not all, AI-powered solutions are created by US-based companies which means they may not be trained on diverse data sets representative of the location they are being used. Just off the top of our head, an AI bot trained in the US may be confused by something as simple as a date format. Even just spelling colour correctly when applying for a copywriter job may get your CV kicked to the kerb.
The other problem worth mentioning is that while AI bots can draw from an enormous corpus of data, it’s very narrow in its pursuit of providing an answer to a question.
Unless the prompts used by a hiring company account for every variable a user may provide, the bot is a failure. Given how diverse people are in their writing styles and CV creation, this is a major problem and we’re doubtful – but happy to be corrected – that hiring companies, recruiters and others are taking the time to properly configure their AI tools.
These tools are also being used in secret. To be clear, we don’t know for certain that the company that sent the message above are using AI but the speed with which a decision was made makes it more likely that a bot made that decision than a human.
In New York City, a law was introduced last year that forces companies to declare when they are using AI.
CBS News reported at the time that companies in the city would also have to conduct regular bias audits “in order to make public the ways in which the AI could be discriminating against certain types of candidates”.
However, New York City is the exception not the rule. AI’s development outpaces the rate at which lawmakers react and development isn’t slowing down at all. As such, by the time local regulators cotton-on to the fact that AI is screening candidates without their knowledge, it may be too late for laws to change those habits.
The war without a winner
For job hunters, AI guarding the gates is a problem especially because they would have no idea what said AI is judging them on. Did their CV’s grammar not follow the AP Style Guide to the letter? Was it the use of colour instead of color that tripped them up? Save for asking the company, the candidate would never know.
For a company the potential risk to their reputation could be a major problem if folks start talking about their job-hunting at a company is going.
Companies also risk introducing bias that may not have previously been present sans AI. Depending how the AI scanner is set up, there could also be a lack of context as to why candidates were rejected.
Was a highly qualified candidate rejected in favour of a candidate with more work experience in an unrelated field? HR would likely have a hard time figuring that out.
But the worst outcome of all is one where companies and job seekers are just using AI to try and one-up one another. AI detection tools are poor for the most part and prone to false positives. There are already hundreds of AI services claiming they can hone your CV with little more than a Google search.
The situation is already spinning out of control and we wouldn’t be surprised to learn that AI scanning of resumes is being held back by a deluge of AI-crafted CVs.
The irony in all of this is that in the job application my friend filled out, the company made sure to state that candidates shouldn’t use AI to draft their cover-letter or their CV.
You first, recruiters.