Men Can’t Stop Faking How Much They Know About AI

Say you’re a guy who reads WIRED—you know a thing or two about artificial intelligence. You recognize the names of well-known people developing AI, like OpenAI CEO Sam Altman. But do you know who Mira Murati is? What about Susan Milner? Craig Dalton?

In a new survey of more than 5,400 Americans, 73 percent of male respondents told Harvard researchers they knew who Altman was. A little over half—53 percent—said they knew of Murati, Milner, and Dalton.

There’s one problem. Murati is a real person: she’s the former CTO of OpenAI and the current CEO of Thinking Machine Labs. Milner and Dalton, on the other hand, do not exist.

Compared to women, “we’re seeing that men are more optimistic and more confident” across the board when it comes to AI, Harvard Kennedy School’s Shorenstein Center executive director Laura Manley tells WIRED. Sometimes, that veers into overconfidence—like claiming to recognize fictional AI executives. (For context, 40 percent of women claimed to know who Milner and Dalton were; 38 percent said they recognized Murati.)

While Manley notes that men are more overconfident in general when surveyed, the report, published on Wednesday, digs into what the confidence gap on AI looks like, as well as the repercussions of gendered divides.

Researchers found that men use AI more frequently and are, on average, better able to correctly answer factual questions about AI. But men and women answer questions incorrectly at around the same rate. The discrepancy is based on how often women said they weren’t sure of an answer—an average of 3.3 questions out of 12 compared to 2.3 questions for men.

Researchers wondered if this predilection toward certainty colors coverage of AI. Across a year of media coverage, men were named in headlines about AI more than 14 times as often as women. Because top AI executives skew heavily male, researchers decided to reexamine coverage, looking solely at people who were quoted as subject experts. Three-quarters of these experts were men—a higher proportion than the actual AI talent pool, which is roughly 29 percent women.

“Maybe there should be more women prominently featured as experts or given the platform to express: not sure,” Manley says.

Besides intellectual bluster, Americans were generally in agreement that AI was freaking them out—men and women both rated “concern” as their top emotional response. However, female respondents were more concerned, fearful, and worried than their male counterparts. Men, meanwhile, rated every positive feeling—including curiosity, enthusiasm, and excitement—higher than women did.

In the current discourse, “there’s so much confidence in either doom or abundance,” Manley says. “It just drowns out all of the voices that should be saying ‘not sure.’” There are many unanswered questions on proper uses for and the future of AI. One gender is simply less likely to acknowledge such uncertainty.

“We have to be able to be able to say ‘not sure’ more often,” Manley says.