Faced with uncertainty, women are more skeptical of AI than men, new research finds
The findings add to a growing literature about gender differences in tech attitudes, which carry implications for tech policy, AI adoption and the future of work, according to Beatrice Magistro.
Women are more skeptical of artificial intelligence and its adoption in the workplace than men — but only when faced with uncertain economic outcomes from the tech, according to new Northeastern University research published in the academic journal PNAS Nexus.
Drawing on survey data from the U.S. and Canada, the study finds that women consistently perceive AI as riskier, especially when its economic effects are uncertain.
The findings add to a growing literature about gender differences in tech attitudes, which carry implications for tech policy, AI adoption and the future of work, said Beatrice Magistro, an assistant professor of AI governance at Northeastern University and co-author of the research.
“We’re not arguing whether AI is good or bad, we’re simply trying to show that this is happening whether we want it or not,” Magistro told Northeastern Global News.
The researchers surveyed roughly 3,000 Canadians and Americans, identifying two key drivers behind the gender gap they uncovered: the notion that women in general are more “risk-averse” than men, Magistro said, and the level of exposure to the potential harms of AI, such as job displacement, bias or widening inequality.
Researchers first asked respondents whether they believed the risks of generative AI outweighed its benefits. They then measured risk in two ways, beginning with what Magistro described as “general risk orientation,” or the respondents’ baseline appetite for risk-taking more broadly.
To measure this, she said researchers posed a lottery-style question to participants — a common approach in economics and behavioral research — to measure an individual’s risk preferences. For instance, respondents were asked to choose between a guaranteed $1,000 payout and a 50% chance of winning $2,000, with a 50% chance of winning nothing. Those selecting the latter, or the “probabilistic option,” were classified as risk-takers, while those choosing the former, or the guaranteed payout, were classified as risk-averse.
Researchers asked respondents to rate on a 0-10 scale whether AI’s risks outweighed its benefits, with the results showing that women were about 11% more likely than men to perceive AI’s risks as outweighing the benefits. That gap is comparable in magnitude, Magistro said, to well-documented gender differences in attitudes toward trade, which researchers used as a benchmark in establishing a baseline for comparison.
“We use trade because there is a long-established literature studying why women are more protectionist than men, and risk is one of the factors that explains that gap,” she said.
Editor’s Picks
When asked open-ended responses about AI’s risks and benefits, women were more likely than men to express uncertainty and skepticism. Women were 6-7 percentage points more likely to say they “don’t know” what the benefits of AI are, and 2-3 percentage points more likely to say that they are “not sure” or “there are no benefits” than men.
They were also more likely to say they did not know what AI’s impact on society writ large and question its benefits, particularly by expressing doubt that the technology would deliver meaningful economic gains, Magistro said.
But Magistro points to a caveat in the data: namely, that the gender gap in AI attitudes hinges on uncertainty. When AI-driven job gains were guaranteed, men and women responded similarly, but when employment outcomes were uncertain, women became significantly more skeptical.
“Basically, when women are certain about the employment effects, the gender gap in support for AI disappears,” she said. “So it really seems to be about aversion to uncertainty.”
Magistro said she thinks the disparity suggests deeper, entrenched patterns of thought and behavior.
“I think part of it has to do with exclusion and socialization,” Magistro said. “Women have historically been more excluded from STEM fields, which have largely been male-dominated. We also have evidence that boys are socialized to be more oriented toward math and science, while girls are less so.”
Indeed, men continue to dominate positions of power in technology firms, while women have had fewer opportunities for exposure to emerging technologies, she said. Data from Women in Tech, a global women’s advocacy organization, in 2025 shows that women make up roughly a quarter of the global tech workforce, and less than a fifth of senior leadership roles.
“Women’s under-representation in science, technology, engineering and mathematics fields means they may have less access to high-paying AI-related jobs and leadership positions,” the researchers write. “As a result, advances in AI could widen existing gender pay gaps.”
Magistro said that the gap between men and women’s attitudes toward AI risk not only speaks to concerns about gender inequality in those scientific fields, but also helps experts track public support for these technologies and the policies that govern them. She said it is important that policymakers address “gender-specific concerns in AI policy,” adding that there is a risk that AI could exacerbate existing inequalities, or “even lead to political backlash” against the tech itself.
