An AI teammate with a female voice boots women’s participation and productivity at work, according to a new study.
Researchers at Cornell University in the US assigned 180 men and women to groups of three and asked them to collaborate virtually on tasks.
A fourth teammate which was an AI agent with either a male or female voice, spoke to them and gave them instructions. Participants then became either a minority or majority member of the team based on gender.
Angel Hsing-Chi Hwang, a postdoctoral associate in information science and the lead author of the study, was feeding lines generated by ChatGPT into the bot that the participants saw.
“Hiring a new person to fill out a team in real-time isn’t realistic,” Hwang said in a statement.
“Our thought was: What if we had on-demand AI agents that can participate and hopefully change team dynamics in a good way?”
All the teams had two tasks: one brainstorming session on conservation ideas and another where they had to review profiles for a job opening and decide who to promote.
The chatlogs of the team conversations were analysed to see how often each person offered ideas or arguments.
The participants in the study also filled out two post-task surveys to rate the level of support offered, the experience on the team and whether they felt marginalised either by a group member or the bot.
Hwang found women in the minority of a particular group participated more when the AI’s voice was female and they reported more positive feelings towards the AI teammate when in the majority.
“Women tended to be pretty quiet when they were the only female members…but when there was a same-gender agent on the team, they contributed a lot more ideas in team discussions,” Hwang said.
When men were in the minority with a male-sounding bot, they were more talkative but less focused on the tasks, the study continued.
“With only a gendered voice, the AI agent can provide a small degree of support to women minority members in a group,” Hwang said.
“We often feel more at ease, and thus work better on teams with people who are like us”.
Hwang said their research opens up questions for further studies into how workplaces can support workers with a long history of being marginalised through the creation of AI agents.
The study only touches on the “tip of the iceberg” with how gender is perceived by people in the workplace, Hwang continued, and future research could consider looking into how performance is changed when those that do not identify as male or female.
