'More rhetoric than real action' with AI use in HR, says expert

A recent report by LinkedIn found that Singapore workers are the world’s fastest in acquiring AI skills, with the city-state recording the highest share of members to have added AI skills to their profiles.

But the verdict’s out as to whether the new tech is actually being used to its best effect in the workplace — especially in HR, says academic Jayanth Narayanan.

“I think one of the challenges I'm seeing is there is more rhetoric than real action,” says Narayanan, Associate Professor at the National University of Singapore (NUS) Business School.

“The Master's program we run at NUS in HR includes a large component where people are learning about AI and analytics and how they can use these technologies. But even though there is a lot of talk about it and an intention to use these technologies in the workplace, there's a gap between what people say and what actually happens in organisations.”

AI shouldn’t replace human interactions in HR

The Business School at the National University of Singapore was one of the first in the world to incorporate modules specifically relating to AI to its HR courses, says Narayanan.

And while AI can automate some tasks traditionally done by HR departments, it cannot replace the human interactions that are so important to the role, he says.

A recent study conducted by NUS Business School in conjunction with the international Institute for Management and the Hong Kong Polytechnic University questioned over 1000 participants regarding the perceived fairness of using AI in the recruitment process.

“Our findings indicate algorithmic decision-making in the recruitment process is less fair than human-assisted methods,” says Narayanan.

The participants in the study were of different nationalities, comprising candidates who had experienced both successful and unsuccessful outcomes in an AI-enabled hiring process.

“We found job applicants perceived a higher degree of fairness when a human is involved in the screening and hiring decision process compared to a fully algorithmic system,” says Narayanan.

“The disparity in fairness is largely attributed to AI’s inability to identify candidates’ unique characteristics compared to human recruiters, who are better equipped to evaluate qualitative information that makes each candidate distinctive,” he says. “AI-enabled processes can overlook important qualities and potentially screen out good candidates.”

Brand and reputational risks to AI

Apart from the impact this has on people’s lives, the brand and reputational risks of relying solely on AI in the recruitment process can be dangerous, Narayanan says. “Algorithmic bias is a concern.”

“For instance, it could turn out there are flaws in the way the algorithms are actually operating - perhaps it has a gender or racial bias.

“If you use the tool to make recruitment decisions, but it leaves out people of a certain race, that can cause significant damage to your reputation on the job market, and your ability to attract future talent.”

There are other considerations around legality, privacy and sensitive employee data, he says.

“It can be dangerous to decide, purely on the basis of an algorithm, not to promote somebody without looking into all the complexities. It may be they had significantly more challenges in this particular reward cycle that made them underperform, or maybe they actually over-delivered given the constraints. It's very hard for a language model to simply understand the uniqueness of a situation. That is a huge concern.”

AI best used as co-pilot

At this point in the technology’s development, the best policy is to not take away the human element to recruitment but to use AI as a co-pilot, says Narayanan.

“I think we do need a human touch. Call me old-fashioned, but I think that for functions like employee wellness, you can't just bring an app that talks to people on a daily basis and expect them to actually feel that the company cares about them.”

Narayanan does see great uses for AI technology as an analysis tool in the HR function - wherever there is very good quality data; for instance, in compensation and benefits.

“Even in hiring, it can be a very good tool to really drill down a large set of candidates to a smaller set,” he says.

“AI’s also useful where there's a lot of legacy data; for instance, exit interviews. You can bring everything into a digital format, and then mine it with good questions. It could enable employers to see patterns emerging from data, down to departmental or regional differentiations.”

As organisations wake up to the possibility of what AI can do, says Narayanan, “they need to be really mindful of the pros and cons of using technology to perform tasks that humans currently do.”