AI may be rewarding candidates who game the system, Maltese recruiters warn

Researchers from the University of Maryland, National University of Singapore and Ohio State University recently found that AI hiring tools consistently favoured CVs rewritten by artificial intelligence over those written by humans – raising fresh questions about bias, fairness and the growing role of AI in recruitment.

Speaking to MaltaCEOs.mt, Head of Group Marketing and Technology at The Remarkable Collective Jonathan Cremona and Konnekt CEO Lara Camilleri offered contrasting but overlapping perspectives on the issue, with both agreeing that human oversight remains essential even as AI tools become more widespread.

Mr Cremona struck a notably optimistic tone, arguing that identifying bias within AI systems is part of the solution itself.

“It is good that we know about this bias, because in knowing so I can create agents that will act as bias police and will keep in check the process,” he said when asked whether the research concerned him from a recruitment perspective.

He described AI recruitment systems as being in their infancy, adding: “The technology is evolving fast and it’s at inception stages, therefore as business leaders and as an agentic champion I feel it is our/my duty to always learn, adapt and improve these systems.”

Ms Camilleri, however, appeared more cautious about the implications of what the research identified as a “Self-Preference Bias.”

“The research is concerning because it identifies a ‘Self-Preference Bias’,” she said, warning that AI systems could end up rewarding candidates who are simply best at using the same AI tools employers themselves rely on.

“There is certainly a risk of shortlisting (rather than hiring) people who are simply best at using the specific AI tools we bought, rather than the best talent,” she added. “This might filter out authentic applicants whose natural style doesn’t align with a model’s training data.”

While both interviewees agreed that AI adoption in recruitment is accelerating, they framed the trend differently.

Mr Cremona argued that all businesses should embrace the technology, particularly sectors already known for rapid tech adoption.

“From a purely business perspective, every business should adopt the technology,” he said. “I am sure that sectors that are more prone to be early adopters of new technologies such as iGaming firms are leading the way.”

He also criticised what he sees as a shallow understanding of generative AI among many businesses.

“People look at LLM systems such as ChatGPT as a glorified search engine or to make their email writing skills marginally better, but this is like you’ve been given a Formula 1 car and you’re driving it like you would drive a Fiat 500,” he remarked.

Ms Camilleri similarly identified iGaming and fintech as likely leaders in AI-assisted recruitment adoption, alongside high-volume industries such as hospitality. However, she stressed that adoption is still relatively early in Malta.

“Usage is still in the early stages, but it will grow rapidly as software providers integrate AI functionality into standard recruitment platforms,” she said. “It’s a provider-led adoption – as the tools become standard, so will their use in Malta.”

Both also confirmed that their organisations already use AI within recruitment-related processes, though neither views the technology as a replacement for human decision-making.

Mr Cremona confirmed that his organisation has implemented AI “throughout different business touchpoints,” including recruitment functions.

Meanwhile, Ms Camilleri explained that Konnekt uses “statistical models for matching roles to candidates” and AI tools to speed up administrative tasks, but emphasised that “a human always makes the final decision on whether to interview.”

“We view AI as a filter for relevance, not a final judge of character and suitability,” she said.

The pair also converged on the idea that the “human in the loop” remains critical, although Ms Camilleri warned that recruiters can still be unconsciously swayed by AI-generated recommendations.

“If a machine gives a ‘Gold Star,’ the human brain is naturally primed to look for reasons to justify that rating,” she said.

She compared this to other forms of bias in recruitment, such as favouring candidates based on appearance or prestigious universities, arguing that training is needed to help recruiters recognise these tendencies.

Mr Cremona, meanwhile, framed AI primarily as a way to improve efficiency by removing repetitive tasks from employees’ workloads.

“The trick is to maximise the ‘human’ efficiency, and this is achieved by offloading repetitive aspects leading to our employees focusing their attention on more important elements of the business,” he said.

The two differed most sharply on whether candidates who do not use AI to polish their applications may eventually be disadvantaged.

Mr Cremona dismissed the concern, arguing that AI-assisted CV writing is rapidly becoming universal among applicants.

“I don’t see this happening,” he said, adding that “the adoption of LLM-assisted CV writing is being adopted fast by all candidates.”

Ms Camilleri, however, warned that the risk is real – especially where shortlisting itself becomes heavily automated.

“The growing use of AI does risk disadvantaging those who write their own CVs especially when the shortlisting is AI automated,” she said.

Still, she argued that structured interviews can help level the playing field.

“The equaliser between AI-written CVs and human-written CVs are structured interviews which move the focus from how a CV was written to how a candidate actually solves problems,” she said.