Study proves female AI 'teammates' increase female participation

A new Cornell University study finds that artificial intelligence-powered virtual teammates with female voices increase female participation and productivity in male-dominated teams.

According to the researchers, the findings suggest that the gender of an AI's voice could have a positive impact on the dynamics of gender-imbalanced teams, which could be useful in designing bots for human-AI teamwork.

“I didn't expect that adding an AI agent to the team would replicate many of the similar results that other researchers have observed in human-only, gender-imbalanced teams,” said Angel-Shin Chi Huang, a postdoctoral researcher in information science at Cornell University's Ann S. Bowers School of Computing and Information Sciences. “That was a surprise to me.”

Fan is lead author of “Sounds of Support: Gendered Voice Agents as Support for Minority Teammates in Gender-Imbalanced Teams,” which received an honorable mention for best paper at the Association for Computing Machinery (ACM) CHI Conference on Human Factors in Computing Systems, held May 11-16. The research was published in the journal Science. Proceedings of the CHI Conference on Human Factors in Computing Systems..

The findings echo previous research in psychology and organizational behavior that shows minority teammates are more likely to join if their team includes people who are similar to them, Hwang said.

“But it's not practical to hire new talent to replenish your team in real time,” said Fan, who studies how AI will affect the way work gets done. “We thought: What if we had an AI agent that could join on demand and change the dynamics of a team for the better?”

To better understand how AI might benefit gender-imbalanced teams, Hwang and Andrea Stevenson-Wong, an associate professor of communication in the College of Agriculture and Life Sciences and a co-author on the paper, conducted an experiment in which they divided about 180 men and women into groups of three and asked them to collaborate virtually on a series of tasks. (The study only included participants who self-identified as either male or female.)

Each group included one woman or one man and a fourth agent with an abstract shape and male or female voice that appeared on the screen to read instructions, suggest ideas and manage time. But the problem was that the bots weren't fully automated. In this experiment, dubbed the “Wizard of Oz” of human-computer interaction, Huang was feeding the bot lines generated by ChatGPT behind the scenes.

After the experiment, Fan and Wong analyzed the chat logs of the team conversations to see how often participants exchanged ideas and discussions. They also asked participants to reflect on the level of support they were provided, their team experience, and whether they personally felt alienated by their human teammates or the bot.

“When we observed the actual behavior of the participants, we saw differences between men and women, and how they responded when there was a female agent on the team versus a male agent,” she said.

“What's interesting about this study is that most participants expressed no preference for a male or female voice,” Wong said. “This suggests that people's social inferences about AI can be influential even when they don't think it's important.”

The researchers found that when women were in the minority, women participated more when the AI ​​had a female voice, whereas men were more talkative but less engaged when working with a male-voiced bot. Unlike men, women had significantly more positive perceptions of their AI teammates when they were in the minority, the researchers said.

“AI agents that have a gendered voice can provide some support to female minority members of a group,” said Fan, who will become a professor at the University of Southern California's Annenberg School for Communications and Journalism this fall.

As for why, Fan points to existing research on team dynamics: “We often feel more secure and perform better in teams with people who are similar to us,” she says.