Virtual-office study finds participants gave more to male-presenting AI

In a virtual-reality office experiment, participants allocated 10.25 percent more money to a male-presenting AI assistant than to a female-presenting counterpart. The assistants used the same underlying technology, according to research by University of Limerick researchers.

The findings indicate that an assistant’s presentation may influence how people respond to it, even when they say gender is unimportant. Participants also judged the male-presenting agent to be more human-like than the female-presenting agent.

Study co-author Mary Hausfeld said: “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.”

How the experiment worked

The researchers had 189 participants carry out tasks in a virtual-reality office alongside AI assistants. These included a text chatbot, a desk robot and human-like agents, which were presented as male or female.

Following the tasks, participants were given money and asked to share it between themselves and the assistants in recognition of the work. The exercise measured how they divided money in the experiment, not decisions about pay in an actual workplace.

What the findings show

Participants said an assistant’s gender did not matter to them and that they preferred AI that was clearly non-human. However, they allocated more to both human-like and male-presenting assistants than to non-human-like and female-presenting ones.

The findings show a difference between participants’ stated preferences and their choices in this experiment. They do not explain the reasons for those choices or show whether the pattern would hold beyond the virtual office. The study’s publication date and venue were not specified.

This report draws on information from Open Magazine.