Gender bias appears to have followed humans into the world of artificial intelligence. A new study has found that female-presenting AI assistants were rewarded significantly less than their male counterparts for performing identical tasks, despite having exactly the same capabilities.
As businesses increasingly introduce AI assistants into everyday operations, the study highlights the importance of designing these systems carefully to avoid reproducing workplace discrimination. Because they operate using technology, (Also read: Indian-origin techie quits $300,000 Meta job to launch AI startup: ‘Wanted to work on something that…’ ) The researchers emphasised that AI systems are not necessarily free from human prejudices simply.
(Also read: ‘AI has changed what I loved’: Software engineer says AI is making coding less meaningful ) According to the New York Post , researchers discovered that participants paid a female AI assistant 10.25% less than a male assistant, raising concerns about how workplace discrimination could extend to emerging technologies. Researchers placed 189 employees in a virtual reality office and asked them to complete workplace tasks alongside different AI assistants. On average, Johanna received 10.25% less money than Johan for completing the same tasks.
The findings revealed a noticeable difference in how participants rewarded the two humanoid assistants. These included a text-based chatbot, a desk robot and two human-like AI agents named Johan and Johanna, designed to appear male and female respectively. Both humanoid assistants had identical programming and capabilities, meaning neither had an advantage in performing the assigned tasks. Participants were given real money to divide between themselves and their AI assistants as compensation for completing the work. The experiment aimed to understand whether an AI assistant’s appearance influenced how people valued its contribution. Participants also perceived the male-presenting assistant as more human-like. Interestingly, several participants who claimed gender did not influence their decisions still ended up rewarding Johan more generously. The researchers also found that participants generally preferred interacting with human-like AI assistants over less human-like alternatives. The findings suggest that gender stereotypes can influence people’s interactions even when they are dealing with machines rather than humans. Dr Hausfeld warned that the way AI assistants are designed could unintentionally reinforce existing social inequalities.
“The technology behind these AI agents was exactly the same, but people did not treat them in the same way,” said Dr Mary Hausfeld of the University of Limerick’s Kemmy Business School in Ireland. “As AI agents become more common in the workplace, we need to think carefully about the characteristics we give them and the behaviors those choices may encourage,” she explained.
