Study Finds Men Are More Likely to Overstate AI Knowledge—and Confidence

A new Harvard-backed survey has found that American men are more likely than women to say they know artificial intelligence experts they have never heard of, underscoring a broader pattern of overconfidence around AI. The research matters because it suggests gender gaps are shaping not only who uses AI, but also who feels comfortable speaking with certainty about it in public debate.

The study, which surveyed more than 5,400 Americans, found that men were significantly more likely to claim familiarity with real and fictional AI figures, while also expressing higher levels of optimism and enthusiasm about the technology. Researchers say the findings point to a confidence imbalance that can distort media coverage, workplace conversations, and policy discussions at a time when AI is rapidly reshaping daily life.

What the Harvard survey found

The clearest signal in the research was simple: men were more likely than women to say they recognized AI names, even when some of those names belonged to people who do not exist. That gap suggests the issue is not just access to information or familiarity with the field, but a stronger willingness among men to answer with certainty.

Researchers asked respondents about several figures associated with AI, including OpenAI co-founder and former chief technology officer Mira Murati. They also included names that were invented for the survey, such as Susan Milner and Craig Dalton, to test whether people would overclaim knowledge.

How often did men say they knew the AI names?

Men were more likely to say they recognized every one of the names presented, including the fictional ones. That included a high level of familiarity with OpenAI chief executive Sam Altman, but also surprisingly strong claims about non-existent AI executives.

According to the survey, 73 percent of male respondents said they knew who Altman was. A little more than half of men, 53 percent, said they recognized Murati, Milner, and Dalton. By contrast, 38 percent of women said they recognized Murati, while 40 percent said they knew the fictional Milner and Dalton.

Harvard Kennedy School’s Laura Manley said the results fit a broader pattern in which men tend to sound more certain in surveys, while women are more likely to admit when they are unsure. She argued that the AI debate needs more voices willing to say they do not have all the answers.

The fabricated names were the sharpest proof point. Murati is real; Milner and Dalton are not. Yet the survey shows how confidence can create the illusion of expertise, especially in a field where many people follow headlines without closely tracking the industry.

Why the confidence gap matters

The confidence gap matters because AI is now being debated in boardrooms, classrooms, newsrooms, and government offices, often by people who are asked to make decisions despite incomplete information. When one group is more willing to answer confidently, even without certainty, it can shape whose voices dominate the conversation.

Laura Manley, executive director of the Shorenstein Center at the Harvard Kennedy School, said the pattern suggests men are generally more optimistic and more confident in AI-related questions. In some cases, she said, that confidence crosses into overconfidence.

That matters for at least three reasons:

• it can elevate people who sound authoritative rather than those who are careful and accurate;
• it can skew media coverage toward more assertive voices;
• it can make uncertainty seem weaker or less legitimate than it really is.

The researchers were especially interested in how a willingness to guess might affect the public understanding of AI. In a fast-moving field where terminology, companies, and technical claims change quickly, overconfidence can spread misinformation without anyone intending to mislead.

How women and men differed on uncertainty

Women were not simply less knowledgeable; the study suggests they were more likely to say when they did not know something. That distinction is important because the researchers found that men and women answered incorrect factual questions at roughly the same rate.

In other words, the main difference was not accuracy so much as certainty.

Participants answered a set of 12 questions about AI. Men, on average, were better at getting factual answers right, but women were more willing to admit uncertainty. Women said they were unsure about an average of 3.3 questions, compared with 2.3 questions for men.

That pattern helps explain how men can appear more knowledgeable even when the underlying performance gap is smaller than it looks. Confidence can function as a social signal, especially in an area like AI where many people are still learning the basics.

What the media landscape shows about AI coverage

The study also looked beyond survey answers and examined how AI is covered in the media. Researchers found that men were named in AI-related headlines more than 14 times as often as women over the course of a year, reflecting the male-heavy concentration of visible AI leadership.

That imbalance becomes even clearer when the researchers narrowed their focus to subject experts quoted in articles. Among those experts, three-quarters were men.

That share is higher than the estimated proportion of women in the broader AI talent pool, which the report puts at roughly 29 percent. The implication is that the media may be reproducing a visibility gap that is already present in the tech industry, while also amplifying it through repeated citation of the same kinds of voices.

Manley said there may be a need for more women to be featured prominently as experts and for public conversations to make room for statements such as “not sure,” rather than rewarding people who appear to have all the answers.

This point is especially relevant in AI coverage, where headlines often revolve around dramatic claims: that the technology will either transform everything or destroy everything. The study suggests that such framing can leave little room for nuance, careful skepticism, or uncertainty.

Why are men more optimistic about AI?

The short answer is that the study does not establish a single cause, but it does show a consistent pattern: men expressed more positive emotions about AI than women did. Men rated curiosity, enthusiasm, and excitement higher, while women reported greater concern, fear, and worry.

At the same time, concern was the dominant emotion for both groups. That means Americans are not neatly split into enthusiasts and skeptics. Instead, many people appear to be holding both fascination and anxiety at once.

What emotions did respondents report?

Both men and women said concern was their strongest feeling about AI overall. But women expressed stronger negative reactions, while men registered stronger positive ones.

• Women: more likely to feel concerned, fearful, and worried.
• Men: more likely to feel curious, enthusiastic, and excited.
• Both: concern remained the top response.

That emotional split matters because feelings often shape how people interpret risk. Someone who is excited about AI may be more inclined to see uncertainty as temporary or manageable. Someone who is worried may be more alert to potential harms and more skeptical of sweeping promises.

How the survey reflects the wider AI debate

The findings arrive at a moment when AI is still being defined in the public imagination. Companies continue to release new models and features, governments are trying to write rules for a technology that evolves faster than policy, and journalists are translating technical developments for general audiences.

In that environment, certainty can be seductive. People who sound confident often get quoted, promoted, and followed, even when the best answer is that the field is still full of unresolved questions.

Harvard’s researchers suggest that this dynamic may be particularly visible among men, who appear more likely to project familiarity and certainty around AI. But the deeper point of the report is broader: the public conversation may reward boldness more than accuracy.

That can be dangerous in a domain where many claims remain contested, including how AI should be regulated, whether it should be deployed in high-stakes settings, and how much its benefits will outweigh its harms.

What does this mean for AI experts and journalists?

It means that both experts and newsrooms may need to rethink who gets treated as authoritative. If the loudest voices are also the most confident, readers may get a distorted picture of where the field actually stands.

The report suggests several practical lessons:

1. give more space to people who can explain uncertainty clearly;
2. avoid relying only on the most visible AI executives;
3. seek a wider range of expert sources, including women;
4. treat certainty claims with healthy skepticism when the evidence is still evolving.

Who are the AI figures at the center of the survey?

The survey used a mix of real and fictional names to test how respondents assessed AI knowledge. One of the central real figures was Mira Murati, a prominent former OpenAI executive who now leads Thinking Machine Labs. Another was Sam Altman, one of the most recognizable names in the industry.

The fictional names, Susan Milner and Craig Dalton, were included to measure whether people would confidently agree that they knew someone they could not actually know. The fact that a sizeable share of respondents did exactly that is what makes the study stand out.

In effect, the research exposed a kind of social shortcut: when a field is unfamiliar but culturally prominent, people may infer that they should recognize its major players, even if they do not. Men appear to have been more willing to take that leap.

What the researchers are really warning about

The warning is not that men know nothing about AI. In fact, the study says men use AI more often and, on average, do better on factual questions. The warning is that confidence can outgrow knowledge, and when it does, it can distort who speaks, who gets believed, and which ideas gain traction.

The authors also seem to be pushing back against two extremes that dominate AI discourse: fatalism and triumphalism. If the only stories told are about disaster or abundance, the middle ground disappears. That middle ground is where most real policy and product decisions have to be made.

Manley said the public conversation has become crowded with voices projecting total confidence about either AI doom or AI abundance, leaving too little room for people willing to say they are uncertain.

That is perhaps the most important takeaway from the research. In a field moving this quickly, uncertainty is not a flaw. It is often the most honest and useful answer available.

Background: why gender and expertise matter in AI

Gender disparities in technology are not new, but AI has given them fresh urgency. Because the field is still young, the visible leaders, quoted experts, and public-facing evangelists are especially influential in shaping the narrative around who belongs in AI and who gets to define it.

If men are more likely to appear in headlines, more likely to be quoted as experts, and more likely to sound certain in surveys, then the cycle reinforces itself. Readers see more men. Editors choose more men. Audiences hear more certainty from men. That creates a feedback loop that can make imbalance appear natural, even when it is socially constructed.

The Harvard study does not solve that problem, but it helps explain one part of it. Visibility is not only about expertise. It is also about confidence, style, and the willingness to answer when others hesitate.

As AI continues to move from niche technology to everyday infrastructure, the stakes of that dynamic are only likely to rise. The people shaping the conversation will matter, but so will the tone they bring to it.

For the moment, the message from this research is plain: in AI, as in many other areas, sounding sure is not the same as being right.

And when the technology itself is still evolving, the ability to say “not sure” may be one of the most valuable forms of expertise of all.