Why we tell Google our in­ner­most thoughts and feel­ings — and that may be a prob­lem for re­searchers

Metro Canada (Vancouver) - - FRONT PAGE - Genna Buck

Hu­mans are a dis­hon­est species.

We’re not hon­est with our bosses, our friends, our fam­ily, our in­ti­mate part­ners, or with re­searchers who phone us for a sur­vey.

But there’s one sit­u­a­tion when we don’t lie. We don’t lie to Google. The search en­gine is si­mul­ta­ne­ously our ther­a­pist, our con­fes­sional and our most trusted friend. Data sci­en­tist and econ­o­mist Seth Stephens-Davi­d­owitz, for­merly of Google, has pored over the un­fath­omably large body of data con­cern­ing when, where and what peo­ple search on Google around the world. In his book Ev­ery­body Lies, he re­ports in­sights about what peo­ple are re­ally think­ing and feel­ing. He’s learned Google users are kinkier than they ad­mit, and more racist too, but most of all, peo­ple are des­per­ately anx­ious and want to know they’re not alone.

You were laughed out of aca­demic jour­nals with this idea. Now ev­ery­body’s in­ter­ested. What hap­pened?

I was a grad­u­ate stu­dent in eco­nom­ics, and I be­came ob­sessed with all you can learn from what peo­ple search, par­tic­u­larly things peo­ple wouldn’t oth­er­wise ad­mit. Peo­ple lie to sur­veys, but they tend to be re­ally, re­ally hon­est on Google. I was con­sid­ered very strange. I couldn’t get the work pub­lished. I couldn’t get an aca­demic job. I was look­ing for other ways to get the work out there, and now I’ve writ­ten this book.

Did you learn any­thing pro­found about hu­man na­ture?

I think anx­i­ety and in­se­cu­rity is prob­a­bly more wide­spread than we usu­ally think. That would be a big take­away. But I think also just peo­ple are com­pli­cated. The tra­di­tional meth­ods of data col­lec­tion have given us a very lim­ited view of the hu­man psy­che.

Do you have a favourite find­ing?

Men make as many searches look­ing for how to give them­selves oral sex as how to give a part­ner oral sex. My other favourite, which you would prob­a­bly also put in the weird or one-off cat­e­gory — but I don’t think it is — is that the top search in In­dia start­ing “My hus­band wants” is “My hus­band wants me to breast­feed him.” It points to this idea that there are facts about hu­man na­ture we didn’t know. There seems to be a some­what wide­spread in­ter­est in adult men be­ing breast­fed in In­dia. It hadn’t been picked up by any of the usual data sources and it isn’t talked about. Well, why In­dia? What causes this to be so much higher in In­dia and nowhere else?

Is this (re­search tech­nique) go­ing to change the way we look at elec­tions? (Searches con­tain­ing ra­cial slurs were strongly associated with ar­eas that un­ex­pect­edly went for Trump, such as Michi­gan and western Penn­syl­va­nia).

Sur­veys are get­ting worse and worse. The re­sponse rate is now un­der 10 per cent. In­ter­net data is get­ting bet­ter and bet­ter. In a cou­ple of elec­tion cy­cles, we’re not go­ing to be us­ing sur­veys any­more. But (with elec­tion pre­dic­tions), we’re just go­ing to pre­dict an event that’s go­ing to hap­pen in three weeks. The at­ten­tion (on elec­tions) is so enor­mous rel­a­tive to its im­por­tance.

If elec­tions aren’t in­ter­est­ing to you, what is?

I talk about child abuse in the book. That’s one area where the data is not good, be­cause most child abuse cases aren’t re­ported. But it turns out a de­cent num­ber of chil­dren, re­ally sadly, make searches such as “my dad hits me” or “my mom beats me.” So we now maybe have the best data ever on when and where child abuse is ris­ing. And I talk about racism and ha­tred. We can break it down minute by minute. We can see how peo­ple are re­spond­ing to the words politi­cians use in speeches — is it calm­ing an an­gry mob, or in­flam­ing an an­gry mob? That is, I think, pretty revo­lu­tion­ary.

What are your find­ings on that front?

There is clearly a strat­egy that is much bet­ter at calm­ing an an­gry mob. And it’s ba­si­cally don’t lec­ture to them, but pro­voke their cu­rios­ity. Talk about Shaquille O’Neal be­ing Mus­lim; don’t talk about how it’s some­one’s re­spon­si­bil­ity not to hate Mus­lims.

What are you look­ing for in the data now?

I’m re­search­ing anx­i­ety. I’ve be­come ob­sessed with it, be­cause there are lot of things in the data that are re­ally, re­ally sur­pris­ing. I’ll give you one ex­am­ple: When Trump was run­ning for pres­i­dent, he was say­ing a lot of scary things. Pretty much all my friends and fam­ily mem­bers and lib­eral peo­ple said they’re ter­ri­fied. Now, if you look at the data in parts of the United States that are re­ally lib­eral, you don’t see an uptick in searches for panic at­tacks or anx­i­ety or any­thing like that.

(Maybe) peo­ple don’t Google anx­i­ety about Trump, even if they’re re­ally anx­ious. I’d be re­ally sur­prised by that. The sec­ond pos­si­bil­ity is peo­ple have a fixed amount of anx­i­ety — they would have been anx­ious about their jobs or their kids, but now they’re anx­ious about Trump. That would be a revo­lu­tion­ary change in how we think about anx­i­ety. The third pos­si­bil­ity is that peo­ple ex­ag­ger­ate how anx­ious they are about Trump be­cause it’s po­lit­i­cally cor­rect, when they tend to ac­tu­ally be much more anx­ious about their own per­sonal sit­u­a­tion. But you don’t bother your friends with that. You sound like a good per­son if you’re anx­ious about Trump.

One of your odd­est find­ings is that peo­ple of­ten type con­fes­sions like “I’m drunk” into Google. What could they pos­si­bly be hop­ing to find?

It’s very strange. It’s a lit­tle bit like the con­fes­sional in Catholi­cism. It is a wide­spread use of Google to type com­plete sen­tences into the search en­gine. You (may) get mes­sage boards where peo­ple are feel­ing sim­i­larly, so you feel less lonely. If you type, “I hate my boss,” you might get mes­sage boards of peo­ple com­plain­ing about their bosses. If you type, “I’m sad,” you get mes­sage boards of peo­ple who are also sad and you re­al­ize, “Oh, I’m not alone.”

Newspapers in English

Newspapers from Canada

© PressReader. All rights reserved.