Confidence about artificial intelligence does not always track knowledge, according to a new Harvard survey reported by WIRED on October 14. In a study of more than 5,400 Americans, researchers found that many respondents claimed to recognize AI figures who did not exist, with men substantially more likely than women to say the fictional names were familiar.

The survey used real and invented names to test respondents’ claimed familiarity with AI leadership. Seventy-three percent of male respondents said they knew who OpenAI chief executive Sam Altman was. Fifty-three percent said they recognized Mira Murati, Susan Milner and Craig Dalton. Murati is a real former OpenAI technology chief who now leads Thinking Machines Lab; Milner and Dalton were fictional.

Women were less likely to claim familiarity across the same test. WIRED reported that 40 percent of female respondents said they knew the two fictional executives, while 38 percent said they recognized Murati. Laura Manley, executive director of Harvard Kennedy School’s Shorenstein Center, told the publication that the pattern fit a wider finding: men expressed more optimism and confidence about AI, sometimes tipping into overconfidence.

Question tiles and anonymous tokens illustrate different patterns of answers and uncertainty.
Men and women had similar incorrect-answer rates, but women more often said they were unsure.

The results were not simply a story of one group knowing more or less. Researchers found that men used AI more frequently and, on average, answered more factual questions about it correctly. Men and women produced incorrect answers at roughly the same rate, however. The main difference was uncertainty: women said they were unsure on an average of 3.3 questions out of 12, compared with 2.3 for men.

That distinction matters because public debate often rewards certainty more than accuracy. A person who offers an answer, even a wrong one, can appear more authoritative than someone who openly admits uncertainty. The survey does not establish that confidence causes influence, but the accompanying media analysis suggests the voices most often elevated in AI coverage are not evenly distributed.

Across one year of coverage examined by the researchers, men appeared in AI-related headlines more than 14 times as often as women. Because many of the industry’s most prominent executives are men, the researchers also examined people quoted specifically as subject-matter experts. Three-quarters were men, a larger share than in the underlying AI talent pool, which WIRED described as roughly 29 percent women.

A media spotlight favors one group of generic expert silhouettes over another.
The researchers found men were quoted as AI experts far more often than women during the period studied.

Men and women did share one broad reaction to AI: concern ranked as the leading emotional response for both groups. Female respondents expressed greater concern, fear and worry, while male respondents rated positive feelings such as curiosity, enthusiasm and excitement more highly. Those responses describe the surveyed population; they do not show how any individual thinks about the technology.

The study’s most useful finding may be less about gender than about the status of uncertainty in technology discourse. Manley argued that confident predictions of catastrophe or abundance can crowd out people willing to say they do not know. In a field with unresolved questions about capability, risk and appropriate use, treating uncertainty as intellectual honesty rather than weakness could improve who is heard and how claims are evaluated.

For organizations making AI decisions, the practical lesson is to separate confidence from evidence. Expertise should be tested through verifiable knowledge, transparent reasoning and a willingness to mark uncertainty. The Harvard findings suggest that without those checks, public discussion can amplify the people most ready to sound certain rather than those most careful about what the evidence supports.