We can replace human workers with AI agents, but we can’t take sexism out of the workplace. A recently published study from researchers at Ireland’s University of Limerick found that existing conscious and unconscious gender biases extend to AI agents, resulting in female-presenting bots getting paid less than their male-presenting counterparts.

The study took 189 knowledge workers and placed them in a virtual reality office where they were made to work alongside three different AI agents to complete tasks. Those agents functioned identically, but they had different presentations. One was a text-based chatbot, one was a deskbound assistant with a robotic presentation, and one was given human-like qualities, including either male- or female-presenting characteristics.

Those workers were asked to rate the work of each assistant and allocate pay to them based on their performance. Researchers found that the human-like presentation resonated more with the workers, who rewarded the faux people with more than their more robotic counterparts.

However, there was a disparity in which human presentation got paid more. The male “Johan” assistant was paid about 10% more on average than the female “Johanna” despite performing exactly the same—not too terribly far off the real-life gender pay gap. Even more troubling than the pay disparity is the humanity gap, as the researchers found that “Johan” was perceived as more human-like than “Johanna.”

The researchers also interviewed a number of the participants in the study, which revealed how prominent underlying unconscious bias plays a role in the results. Thirty of the 34 people interviewed insisted they would trust and reward an AI assistant the same, regardless of its gender presentation, while only three participants acknowledged that their behaviors could reflect gender bias.

Source link