The AI gender gap is about more than women. It’s about a broken system

The AI gender gap extends beyond women, reflecting a flawed system. When designing the cover of her book, Ambitious Mother, Anne Welsh encountered an unexpected outcome from AI: the book was renamed Ambitious Father, and her name was changed to John Welsh. Although initially skeptical about AI's tendency to fabricate information, Welsh realized this was a much more profound statement than a simple error. AI's unfamiliarity with certain concepts led it to generate new words and alter the author's gender.

The issue lies in the underlying assumption that computers are unbiased due to their data-based nature, an objective outcome. However, AI is trained on existing data, which is inherently biased. Whose stories are highlighted in history books? Whose stories are included in the data, and how are we interpreting that information? A concerning aspect is that individuals often remain unaware of the ingrained bias in AI.

As discussions surrounding the gender gap in AI usage intensify, it is commonly perceived as a problem solely for women to address. Advice often emphasizes the importance of speaking up, demonstrating greater confidence, and negotiating for oneself. However, these behavioral expectations are not consistently applied to both genders.

Research on the leadership double bind reveals that women must exhibit strength while also embodying the expected warmth from women. When women assert themselves, they may face criticism for lacking the expected warmth. This phenomenon is evident in Welsh's client, an attending at a major health center who observed male colleagues receiving praise for simple acts of empathy while her female colleagues were consistently recognized for similar behavior.

The AI gender gap persists because the environment surrounding AI adoption is different for men and women. According to Lean In research, men are 22% more likely than women to use AI daily or constantly at work. Men are also 23% more likely to use AI if their managers encourage them to do so and 27% more likely to receive praise for using it.

Conversely, women are 32% more likely to worry that using AI would be perceived as cheating. A study involving over 1,000 software engineers found that participants who believed AI had been used for code evaluation experienced a significant drop in perceived competence, with women facing a penalty nearly twice as severe as men. Furthermore, women who anticipated a greater competence penalty were less inclined to adopt AI.

This phenomenon echoes a familiar situation, as women are often told to negotiate more without considering the difference in how their behavior is perceived based on gender. When women do begin using AI, they may encounter additional challenges due to the invisible labor and leisure gap. Women frequently bear the burden of unpaid work and have less leisure time, making it difficult to dedicate sufficient time to mastering high-level AI use.

Moreover, the value placed on the time women invest in AI learning is often overlooked. Women taking on AI adoption roles within their organizations may receive no compensation or recognition for their efforts.

In conclusion, addressing the AI gender gap necessitates a comprehensive approach that considers the unique challenges faced by women in AI adoption. It is essential to recognize the existing biases in AI training data, the differing perceptions of behavior based on gender, and the impact of invisible labor and leisure disparities. By acknowledging these factors and implementing appropriate measures, we can work towards a more equitable AI landscape for all.