The research conducted by Cornell University found that an artificial intelligence-powered virtual teammate with a female voice can increase participation and productivity among women in teams dominated by men. The study suggests that the gender of an AI’s voice can have a positive impact on the dynamics of gender-imbalanced teams and may influence the design of bots used for human-AI collaboration. This research builds on previous studies that have shown minority teammates are more likely to engage if they are joined by members similar to them. Angel Hsing-Chi Hwang, the lead author of the paper and a postdoctoral associate in information science, conducted the study with Andrea Stevenson Won, an associate professor of communication.
To explore how AI can support gender-imbalanced teams, around 180 men and women were assigned to groups of three and asked to collaborate virtually on a series of tasks in an experiment. Each group had either one woman or one man, and an additional abstract shape with a male or female voice acted as a bot, providing instructions, ideas, and timekeeping. Hwang, using ChatGPT-generated lines, controlled the bot behind the scenes in what is known as a “Wizard of Oz” experiment. The researchers analyzed the chat logs of team conversations to determine how often participants contributed ideas or arguments and also collected feedback from the participants after the experiment.
The results showed that when women were in the minority on a team, they participated more actively when the AI’s voice was female. In contrast, men in the minority were more talkative but less focused on tasks when working with a male-sounding bot. Women reported significantly more positive perceptions of the AI teammate when they were in the minority, while men did not express a preference for a male or female voice. This suggests that people’s social inferences about AI can influence their interactions even if they do not believe the gender of the voice is important.
The research highlights the potential for AI agents with gendered voices to provide support for women who are outnumbered in a group. By incorporating a female voice for the AI teammate, women in gender-imbalanced teams felt more engaged and included in the collaboration process. The study also sheds light on the impact of the AI’s voice on communication patterns and task orientation within teams. These findings could inform the design and development of AI systems used for teamwork and collaboration in various settings, emphasizing the importance of considering gender dynamics in human-AI interactions.
