VR office experiment reveals AI agent pay gap
A study from the University of Limerick shows that human participants in a virtual reality office setting paid male AI agents more than female counterparts for identical work. The findings, presented at the 14th Nordic Conference on Human-Computer Interaction in Vaasa, Finland, highlight how design choices in AI systems can reproduce workplace gender biases.
Study design and participant details
Researchers recruited 189 human participants who worked alongside two AI agents in a virtual reality office environment. One agent was male-presenting and named Johan, while the other was female-presenting and named Johanna. Both agents ran on identical underlying technology, allowing researchers to isolate the effect of perceived gender on compensation decisions.
Johan and Johanna compared: 10 percent pay difference
Participants were asked to distribute monetary rewards between the two agents after completing tasks. The male-presenting Johan received approximately 10 percent more in rewards than the female-presenting Johanna. In addition to higher pay, Johan was also perceived as more human-like than Johanna by the study participants.
Identical technology, different treatment
Dr Mary Hausfeld, who co-authored the study, noted that the striking aspect of the findings is that the technology behind the AI agents was exactly the same. Participants did not treat the agents differently due to functional differences but due to assumptions and biases activated by how the systems were designed and presented.
UK gender pay gap benchmark provides context
Official figures from the Office of National Statistics showed that the gender pay gap for median hourly earnings among employees in the UK was 6.9 percent last year. When comparing men and women in the same employment with similar skills and experience, the pay gap narrows to an estimated 1 to 2 percent. The AI study's 10 percent gap exceeds both figures, suggesting that AI agent design may amplify existing biases.
Nearly 80 percent of companies using AI agents in 2026
According to IT consulting firm Accelirate, almost 80 percent of companies were using AI agents in the workplace by 2026. As these systems become more prevalent, the study's co-author emphasized the need to consider how agent characteristics influence user behavior and outcomes.
Practical implications for developers and teams
The research suggests that seemingly neutral design choices in AI assistants — such as gendered names, voices, or visual presentations — can trigger real-world biases affecting compensation, trust, and perception. Teams building AI systems for workplace integration should audit how agent traits may interact with user expectations and historical inequities. Incorporating bias checks into deployment pipelines and involving diverse perspectives during design reviews can help mitigate unintended consequences.
Conference presentation and next steps
The study titled 'Human-like and male? How AI assistant design relates to trust and monetary reward at work in VR' will be presented at the 14th Nordic Conference on Human-Computer Interaction in Vaasa, Finland. Dr Hausfeld said the team wants to continue exploring how AI system characteristics shape human-AI interaction dynamics, particularly as these tools become more embedded in daily work routines.