Can Female AI Teammates Enhance Women's Workforce Participation?

Is there a connection between the predominance of men in AI training and the disproportionately adverse effects of AI on women's careers? Mercer’s 2024 Global Talent Trends Study dives into this issue, analysing whether women's wariness of AI is justified and how AI's disruptive potential impacts gender dynamics in the workforce.

The Experiment And Its Implications

Researchers Angel Hsing-Chi Hwang and Andrea Stevenson Won conducted a meticulous experiment to explore how the gender of an AI's voice influences team dynamics. The study involved approximately 180 participants, divided into groups of three, collaborating virtually on a series of tasks. Each group included either one man or one woman and a fourth AI agent, represented by an abstract shape with either a male or female voice. The AI agent’s role was to read instructions, contribute ideas, and manage timekeeping. Importantly, this AI was not fully automated; instead, Hwang operated it behind the scenes, feeding it lines generated by ChatGPT, in what is known as a “Wizard of Oz” experiment.

Key Findings

The analysis of team interactions yielded several significant insights

1. Increased Participation of Women: Women in the minority showed higher levels of participation when the AI agent had a female voice. This suggests that the presence of a gender-similar AI can encourage minority members to engage more actively in team discussions.
2. Behavioural Differences in Men: Men who were in the minority also showed increased participation with a male-voiced AI, although their focus on tasks decreased, indicating that while gender-similar AI voices can boost engagement, they may also affect the quality of participation differently for men and women.
3. Perception of AI Teammates: Women reported significantly more positive perceptions of the AI teammate when it had a female voice, especially when they were in the minority. This was not mirrored by the male participants, indicating a gender-specific response to AI voices.

These results are consistent with previous research in psychology and organisational behaviour, which indicates that minority group members are more willing to join when they consider team members similar to themselves.

Gender Disparity In AI Adoption

A report by Mercer's sister company, Oliver Wyman finds a significant gender difference in using generative AI tools. Approximately 59% of men aged 18 to 65 use AI products every week, compared to 51% of women of the same age. This disparity is even more prominent among young workers, with 71% of men aged 18–24 using AI weekly vs 59% of women.

According to Mercer's 2024 Global Talent Trends Study, women would be disproportionately affected because they hold more of the positions that are predicted to be disrupted by AI. For example, the administration, healthcare, education, and social services industries all have strong female representation and are among the sectors most likely to see substantial job losses as a result of AI and automation.

Kate Bravery, Mercer Global Advisory Solutions & Insights Leader, believes that while women are now suffering more disruption as a result of AI, she asks whether this disruption is related to the fact that more men are involved in training AI models.

"Women definitely occupy more jobs that the first wave of generative AI is impacting, but because of the gender imbalance, as generative AI penetrates further up the hierarchy, we'll start to see some of that change," she stated. 

Bravery relates this gap to executives' excitement and support of AI, who are generally more aware of its productivity benefits. The findings indicate that younger generations are not as evenly exposed to AI as one might predict, stressing the need for concentrated efforts to bridge this gap.

Bravery al revealed a statistic from an upcoming survey (Mercer has yet to release the data), which provides insight on generative AI adoption — and it's all about the messenger, "When business executives talk about using AI, it's 50% more likely that people in their organisation also are. But when it's the chief human resources officer talking about using it as the future way of working, it's 70% more likely. That's why I think that the person who leads the people's function is best placed to be championing that."

To mitigate these biases, companies must establish transparency in internal labour flows and monitor gender representation at all levels. Understanding recruiting, retention, and progression patterns can reveal whether women are given equal opportunity. Furthermore, organisations should prioritise three critical dimensions to reduce negative outcomes for women: access to rewards and pay, representation at all organisational levels, and possibilities for career mobility and training.

Practical Applications and Future Directions

The Global Talent Trends study’s results suggest several practical applications for the design of AI systems in workplaces:

• Gender-Sensitive AI Design: AI developers should consider incorporating gender-specific voices into virtual assistants to foster more inclusive and participatory team environments.
• Dynamic Role Allocation: AI systems can be designed to dynamically adapt their roles based on team composition, potentially taking on more supportive or directive roles as needed to balance team dynamics.
• Feedback Mechanisms: Implementing feedback mechanisms within AI systems to monitor team interactions and adjust responses ireal timeme could further enhance their effectiveness.

The results of this study indicate a future in which AI can be tuned to address specific social dynamics, hence increasing inclusivity and productivity across various domains. AI's ability to support minority group members in professional contexts may result in broader societal benefits, such as enhanced workplace diversity and equality.