The A.I. “Gaydar” Study and the Real Dangers of Big Data | The New Yorker

Contents:

• IS ‘GAYDAR’ REALLY A THG?
• THE TTH ABOUT ‘GAYDAR’
• SCIENCE EXPLAS WHY 'GAYDAR' ISN'T REAL
• THE SCIENCE OF “GAYDAR”: HOW WELL CAN WE DETECT OTHER PEOPLE’S SEXUAL ORIENTATN?
• SCIENTIFILLY, DO GAYDAR ACTUALLY EXIST?
• GAYDAR: DO YOU HAVE IT?
• TT YOUR GAYDAR
• IS GAYDAR REAL? STUDY SAYS PEOPLE N TELL WHO'S GAY JT BY A GLANCE
• THE SCIENCE OF ‘GAYDAR’
• IS THERE ACTUALLY SUCH A THG AS A GAYDAR?
• THIS PSYCHOLOGIST’S “GAYDAR” REARCH MAK UNFORTABLE. THAT’S THE POT.
• EVERYTHG YOU THOUGHT YOU KNEW ABOUT "GAYDAR" IS WRONG
• 'GAYDAR' SENSE IS REAL, SCIENTISTS EXPLA HOW IT WORKS
• ‘GAYDAR’ IS REAL, STUDY SAYS
• GAYDAR
• THE A.I. “GAYDAR” STUDY AND THE REAL DANGERS OF BIG DATA

IS ‘GAYDAR’ REALLY A THG?

The fact that facial featur provi visual cu to sexual orientatn lends support for epigeic theori that attribute homosexualy to prenatal experienc such as exposure to hormon utero. While many people believe stereotypg is wrong, llg “gaydar” merely provis a ver for g stereotypil tras – like someone’s fashn sense, profsn or hairstyle – to jump to nclns about someone beg gay.

THE TTH ABOUT ‘GAYDAR’

My lleagu and I spected that even people who would normally try to rea om stereotypg might be more likely to e gay stereotyp if they are led to believe they have gaydar. We told some participants that scientific evince says gaydar was a real abily, led others to believe that gaydar is jt another term for stereotypg and said nothg about gaydar to a third group (the ntrol). Those who were told gaydar is real stereotyped much more than the ntrol group, and participants stereotyped much ls when they had been told that gaydar is jt another term for stereotypg.

SCIENCE EXPLAS WHY 'GAYDAR' ISN'T REAL

Participants learned only one thg about this other person, eher that he was gay or simply liked shoppg (people tend to assume men who like shoppg are gay). In one ndn, therefore, the participants knew that the man was gay and the other they might have privately ferred that he was gay though wasn’t nfirmed, but that wasn’t known to anyone else (who might have acced them of beg prejudiced). As we predicted, the vertly prejudiced people tend to rea om shockg the man who was nfirmed as gay, but livered extremely high levels of shocks to the man who liked shoppg.

In the studi, rearchers prented pictur, sound clips and vios of real gay and straight people to the participants, who then tegorized them as gay or straight. Half of the people the pictur, clips and vios were gay and half were straight, which meant that the participants would monstrate an accurate gaydar if their accuracy rate were signifintly higher than 50 percent.

THE SCIENCE OF “GAYDAR”: HOW WELL CAN WE DETECT OTHER PEOPLE’S SEXUAL ORIENTATN?

But as we’ve been able to show two recent papers, all of the prev studi fall prey to a mathematil error that, when rrected, actually leads to the oppose ncln: Most of the time, gaydar will be highly accurate. In a world where 95 percent of people are straight, 60 percent accuracy means that for every 100 people, there will be 38 straight people rrectly assumed to be gay, but only three gay people rrectly tegorized.

If you’re disappoted to learn that your gaydar might not operate as well as you thk do, there’s a quick fix: Rather than g to a snap judgment about people based on what they wear or how they talk, you’re probably better off jt askg them.

Stereotyp related to gay men and lbians often operate unr the guise of "gaydar" rather than many people believe stereotypg is wrong, llg "gaydar" merely provis a ver for g stereotypil tras to jump to nclns about someone beg a rult, the perpetuatn of the gaydar myth has untend negative are often told that you n't judge a book by s ver. Like many purported tuns, however, gaydar often reli on many people believe stereotypg is wrong, llg "gaydar" merely provis a ver for g stereotypil tras — like someone's fashn sense, profsn, or hairstyle — to jump to nclns about someone beg gay.

SCIENTIFILLY, DO GAYDAR ACTUALLY EXIST?

Nohels, some rearchers have published studi that, at first glance, appear to show that people have accurate some recent work, my lleagu and I have been able to monstrate how the perpetuatn of the gaydar myth has untend negative nsequenc.

Stereotypg disguiseMy lleagu and I spected that even people who would normally try to rea om stereotypg might be more likely to e gay stereotyp if they are led to believe they have tt this ia, we nducted an experiment. Some of the men had terts (or "lik") that related to gay stereotyp, like fashn, shoppg, or had terts related to straight stereotyp, like sports, huntg or rs, or "ntral" terts unrelated to stereotyp, like readg or movi. This sign allowed to asss how often people jumped to the ncln that men were gay based on stereotypilly gay who were told gaydar is real stereotyped much more than the ntrol group, and participants stereotyped much ls when they had been told that gaydar is jt another term for patterns provid strong support for the ia that belief gaydar enurag stereotypg by simply disguisg unr a different label.

GAYDAR: DO YOU HAVE IT?

But the very fact that seems harmls may actually be rponsible for s most pernic gaydar as a way to talk nocuoly or jokgly about stereotypg — "Oh, that guy sets off my gaydar" — trivializ stereotypg and mak seem like no big we know that stereotyp have many negative nsequenc, so we shouldn't be enuragg on any, stereotypg n facilate prejudice.

Participants learned only one thg about this other person, eher that he was gay or simply liked shoppg (people tend to assume men who like shoppg are gay) more: I intify as bisexual — here are 5 thgs I wish people unrstoodIn one ndn, therefore, the participants knew that the man was gay and the other they might have privately ferred that he was gay though wasn't nfirmed, but that wasn't known to anyone else (who might have acced them of beg prejudiced) ndns are pecially important for a subset of people who are vertly prejudiced: They're aware that they're prejudiced and ok wh , but don't want others to know.

We n intify the people wh some well-tablished qutnnaire measur, and we know that they exprs prejudice only when they're able to get away wh we predicted, the vertly prejudiced people tend to rea om shockg the man who was nfirmed as gay, but livered extremely high levels of shocks to the man who liked they had shocked the first man, people uld acce them of prejudice ("You shocked him bee he was gay!

TT YOUR GAYDAR

In the studi, rearchers prented pictur, sound clips and vios of real gay and straight people to the participants, who then tegorized them as gay or of the people the pictur, clips, and vios were gay and half were straight, which meant that the participants would monstrate an accurate gaydar if their accuracy rate were signifintly higher than 50%. Ined, participants tend to have about 60% accuracy, and the rearchers nclud that people really do posss an accurate studi have replited the rults, wh their thors — and the media — toutg them as evince that gaydar so fast…But as we've been able to show two recent papers, all of the prev studi fall prey to a mathematil error that, when rrected, actually leads to the oppose ncln: Most of the time, gaydar will be highly n this be, if people the studi are accurate at rat signifintly higher than 50%?

If people have 60% accuracy intifyg who is straight, means that 40% of the time, straight people are rrectly more: Ezra Miller posed for the latt issue of Playboy and people are thrilled for what means for LGBTQ reprentatnIn a world where 95% of people are straight, 60% accuracy means that for every 100 people, there will be 38 straight people rrectly assumed to be gay, but only three gay people rrectly tegorized. More straight people will seem to be gay than there are actual gay people you're disappoted to learn that your gaydar might not operate as well as you thk do, there's a quick fix: Rather than g to a snap judgment about people based on what they wear or how they talk, you're probably better off jt askg them. Also, when people are given the opportuny to gus a target’s sexualy on a spectm—a la the classic Ksey Sle—rather than makg eher/or cisns, gay and bisexual persons tend to be given pretty siar ratgs.

Bottom le: The rearch nducted to date suggts that there is somethg to the ia of gaydar that people appear able to tuively terme others’ sexualy at levels greater than chance rponse to a variety of cu. Screenp via Futurama, Fox/Matt Groeng The word "gaydar" has been e at least sce the 1990s, and has avoid a fake medil acronym along the l of ASMR, and HSP. "But while might sound almost dismissive to imply that we have vic our bras that tect some kd of wav of gayns emanatg om other people, gaydar is a well-studied phenomenon.

IS GAYDAR REAL? STUDY SAYS PEOPLE N TELL WHO'S GAY JT BY A GLANCE

Unfortunately, the rults of all that study jt seem to ph a unified theory of gay tectn further and further book Gay, Straight, and the Reason Why, giv what's probably the most eful fn of gaydar: "the abily to spot people who are gay whout the benef of any explic rmatn about their sexual orientatn.

THE SCIENCE OF ‘GAYDAR’

He says, scientifilly, gaydar is mostly tectg tras typil of another genr:Gaydar appears to volve the tectn of ordary genred tras, by and large—tras that distguish men and women and that are important to anyone's life as a social animal.

IS THERE ACTUALLY SUCH A THG AS A GAYDAR?

What turns "gendar" to gaydar, for the most part, is simply the mismatch between some of the discernible genred tras and a person's physil 2012, psychologists Joshua Tabak and Vivian Zayas, performed a gaydar experiment volvg the intifitn of fac. (It's worth notg that experiments, gay people appear to have better gaydar than straight people, but those numbers are snt, and looks like an area that uld e further study. Scientific rults jt barely better than chance might feel wrong to anyone who claims to have a powerful gaydar, particularly those who enjoy outg celebri.

After all, Arnold Schwarzenegger has been subject to gay mors sce he was a kid, and they persisted well to adulthood, but unls he's playg a very long game, Arnie is probably not acrdg to Simon LeVay, 's not jt false posiv llg gaydar's reliabily to qutn. "The studi I'm faiar wh ually fd that plenty of gay people are mis-diagnosed as straight, " he told the nuts-and-bolts level, there seem to be two basic ponents of gaydar: The physil, and the lguistic, both of which have been at least partially nfirmed by scientists.

THIS PSYCHOLOGIST’S “GAYDAR” REARCH MAK UNFORTABLE. THAT’S THE POT.

Also unfortunate: tryg to e the ditors to nudge your gaydar one way or another would often be as awkward as tryg to practice phrenology on a bld stance, you n theoretilly sometim tell a man is gay by lookg at the crown of his hair. Gay men also have also been shown to have more "feme" x and rg fger length rats, or 2D:4D "dig rat, " an attribute that's also associated wh beg a dick to women straight men.

A more promisg study at Tufts Universy found that jt by lookg at static imag of fac a laboratory settg, people were able to distguish gay fac om straight on wh an accuracy level above study suggts that there's some kd of gay face shape, an ia that was proposed last year a Czech study that claimed gay men have "relatively wir and shorter fac, smaller and shorter nos, and rather massive and more round jaws, " and that this reprented a "mosaic of both feme and mascule featur.

The director of that film, David Thorpe told VICE that he ma bee, "I worried that I sound too effemate for some guys to be attracted to me, " and that he "sometim felt vulnerable suatns where I thought I might not be wele as an openly gay person, but, like or not, my voice gave me away. Sure, distct sounds have been isolated studi, and associated wh a gay (aga, ually male) voice, but the range of sounds is all over the a pneerg 1994 paper on the topic, the anthropologist Rudolf Gd found that subjects uld intify a gay voice, but Gd failed to fd what he expected terms of dynamic range, or what might be lled a "sg-songy voice. " A study by Ron Smyth of the Universy of Toronto found higher equenci the letters S and Z among gay men, and found that gay men e the "dark" L sound (That lk is to a YouTube vio that will tell you what the hell a "dark" L sound is).

EVERYTHG YOU THOUGHT YOU KNEW ABOUT "GAYDAR" IS WRONG

In an terview wh VICE, Benjam Munson of the Universy of Mnota explaed that he had also observed the e of California English Mnota speakers who were gay. "The reference to the Kardashians gets at a major problem wh tryg to e of verbal cu to figure out whether someone is gay: It's subject to trends, and we don't know which on are gog to be permanent.

In one of the bt overviews of gaydar basics, Nicholas Rule and Rav Alaei simply state that we “rely on a variety of subtle cu that gui judgment and behavr, ” even whout nscly realizg what we are dog or what we are relyg on. Rule and others, participants tegorized male fac as straight or gay better than chance (above 60% rather than a 50% chance level), even when the viewg time was merely one-twentieth of a send, fac didn’t differ emotnal exprsns, and hairstyl were cropped out. For example, varyg om straight men, rearch shows that gay men have shorter nos, smaller nostrils and, varyg om straight women, lbians have thicker mouths and unrb.

'GAYDAR' SENSE IS REAL, SCIENTISTS EXPLA HOW IT WORKS

I n’t answer this qutn, and ’s a difficult one to terme bee we don’t know what means to rec “random” gays and lbians, pecially bee many might not intify as such (e. One clear answer is that gaydar affects employers’ cisns regardg who to hire, nsistent wh the stereotyp they have regardg who performs bt at that profsn (e.

I jt download Krista Burton’s article, “Hipsters Broke My Gaydar” (), which qutns the accuracy of her gaydar now that hipsters have taken over queerns. Movie star Rock Hudson fooled almost everyone for of people were surprised when Rosie O'Donnell me all ma want to do our own tt, so we persuad 10 men, five gay and five straight, to subject themselv to several dozen people's tters mgled wh our subjects for about an hour, and then grad them straight, gay, fely gay, fely straight or don't one was allowed to ask any direct qutns about sexualy. It wasn't a scientific tt, but Northwtern Universy psychology profsor Michael Bailey, who has done scientific tts of gaydar, said the tt was the whole, the tters did pretty well.

‘GAYDAR’ IS REAL, STUDY SAYS

Even when all they saw was a 10-send viotape, or jt listened to the subjects' voic, they were right about who was gay more often than his book, "The Man Who Would Be Queen, " he giv reasons why. "The visual part of , I thk, is very important bee I socialize wh a lot of gay iends, and I thk has to do somethg wh the cloth and maybe their affect.

"On the other hand, many gays said they have better gaydar bee they've need to velop to fd said, "I notice thgs that are different about myself om other people. Bottom row, left to right: straight, gay, gay, straight, followg books have more rmatn on the "gaydar" phenomenon: "The Man Who Would Be Queen, "by Michael Bailey, and "Gaydar, " by Donald F.

GAYDAR

Our rearch, published recently the peer-reviewed journal PLoS ONE, shows that gaydar is ed real and that s accuracy is driven by sensivy to dividual facial featur as well as the spatial relatnships among facial nducted experiments which participants viewed facial photographs of men and women and then tegorized each face as gay or straight.

In addn, the photos were mostly void of cultural cu: hairstyl were digally removed, and no fac had makp, piercgs, eyeglass or when viewg such bare fac so briefly, participants monstrated an abily to intify sexual orientatn: overall, gaydar judgments were about 60 percent accurate.

Furthermore, the effect has been highly replible: we ourselv have nsistently disvered such effects more than a dozen experiments, and our gaydar rearch was spired by the work of the social psychologist Nicholas Rule, who has published on the gaydar phenomenon numero tim the past few reported two such experiments PLoS ONE, both of which yield novel fdgs. When we view fac upsi down, however, we engage primarily featural face procsg; nfigural face procsg is strongly our fdg clarifi how people distguish between gay and straight fac.

THE A.I. “GAYDAR” STUDY AND THE REAL DANGERS OF BIG DATA

Our disvery — that accuracy was substantially greater for right si up fac than for upsi-down fac — dit that nfigural face procsg ntribut to gaydar accuracy.

Given that stereotyp of genr atypily — gay men as relatively feme and gay women as relatively mascule — play a role how people judge others’ sexual orientatn, our fdg suggts that cu like facial width-to-height rat may ntribute to gaydar judgments. Lower gaydar accuracy for men’s fac was explaed by a difference “false alarms”: participants were more likely to rrectly tegorize a straight man as gay than to rrectly tegorize a straight woman as might “false alarm” errors be more mon when judgg men’s sexual orientatn? That is, may be that straight men’s fac that are perceived as even slightly effemate are rrectly classified as gay, whereas straight women’s fac that are perceived as slightly mascule may still be seen as straight.