Even Male AI Bots Get 'Paid' More Than Their Female Counterparts

When can we say that AI has true human-like intelligence, rather than just impressive coding? Is it when large language models (LLMs) feel hunger, pain, or fear? When bots develop a worldview? 

I’ll be honest – answering that is above my pay grade. Speaking of which, AI agents have gotten closer to people in one regard: they just got their very own wage gap. 

Yes, you read that right. New research from the University of Limerick has shown that even when the “women” we’re working with aren’t real, we’re still inclined to pay them less than the “men”.

Real-life influences seem to map onto AI agents

In the new study, researchers asked 189 participants to complete work-related jobs with human-like AI bots in a virtual office. 

One of these digital agents looked and sounded male (Johan), while another presented as a woman (Johanna). 

People taking part in the study were given real money, which they were asked to divide between themselves and their on-screen co-worker. This was meant to show what they saw the assistant’s efforts as being “worth”. 

Quelle surprise: those who’d worked with Johanna typically gave less of the cash away than participants who’d been paired with Johan. In all, she earned 10.25% less than he did for the same work.

This, despite the fact that both of the “assistants” were made from the same technology. In other words, there was no reason – and probably no way – for them to behave differently enough from one another to merit this discrepancy.

University of Limerick’s Dr Mary Hausfeld, who co-authored this study, said: “What is striking about our findings is that the technology behind these AI agents was exactly the same, but people did not treat them in the same way.”

AI, which is made by humans, may not be neutral 

“We often think of AI as being neutral, but the way we design and present these systems can activate those same assumptions and biases that exist in our interactions with other people,” Dr Hausfeld continued. 

We have already seen AI propogate existing racial bias in clinical models, for instance.

“As AI agents become more common in the workplace, we need to think carefully about the characteristics we give them and the behaviours those choices may encourage,” the study’s co-author added.

“We don’t want to inadvertently reproduce existing inequalities in a new technological setting”.