What Google Searches Reveal About Who We Really Are (It’s As Weird As You Think)

“You can’t just ask people what they think. You can’t just ask people who they are. They reveal it.”

READ ON TO DISCOVER:

  • How Google search data could have predicted Trump’s election
  • The truth about America’s racial divide
  • Which country has a highly unusual sexual preference

Seth Stephens-Davidowitz is a visiting lecturer at the Wharton School and a former Google data scientist who received his PhD in economics from Harvard. Derek Thompson is a senior editor at The Atlantic and author of Hit Makers, which explores the secret histories of pop culture hits and the science of popularity. The two recently sat down to discuss the counterintuitive, often downright bizarre revelations in Seth’s new book, Everybody Lies, in which he examines vast quantities of Google search data to shed new light on our fears, our sexual preferences, and what it means to be human.

This conversation has been edited and condensed. To listen to the full version, click the SoundCloud link below.

Derek: You are the master of gleaning our secret psychology from what we search on the internet. What are the positive and negative, light and dark things about ourselves that we can see?

Seth: Well, the thesis of the book is that you can’t really trust what people tell you in everyday life, or even what they tell surveys. But for some reason, people confess their secrets to Google, and we can learn for the first time who we really are.

Derek: And who we really are turns out to be both good and bad. On the good side, you recognize that even the strangest things that you feel aren’t actually that weird, that our weirdnesses tend to be shared. Yet sometimes what we reveal on Google makes us seem like terrible people.

Seth: The dark side is a big theme of the book. I start with talking about racism. If African Americans in this country thought that there was a huge number of Americans who, while saying nice things to their faces, were secretly going on the internet searching for racist jokes, and not voting for black people when they ran for office, that’s 100% true. It’s shown in the data.

Derek: When you say “shown in the data,” what are you looking at, and where can you see it?

Seth: I looked at searches that include the N-word. I thought it would be searches for rap lyrics, [but] it’s not rap lyrics, or the version [of the word] that ends in ‘a.’ It’s searches for jokes mocking African Americans. In the time period I was studying, they were made about as frequently as searches for “Lakers” and “migraine” and “economist” and “Daily Show.”

And they were made in places of the country [where] I didn’t think racism was that high. You think of the country’s history, it’s natural that you think racism would be highest in the South. And it definitely is. But right up there with those states are upstate New York, western Pennsylvania, eastern Ohio, industrial Michigan, and rural Illinois. The real divide in racism these days is not North versus South, it’s East versus West. It’s much higher east of the Mississippi River than west of the Mississippi River.

Derek: You were finishing this book in the middle of the presidential election, and there were a lot of articles showing that support for Trump seemed to correlate with racist searches. Did you find any information that corroborated that thesis?

Seth: Nate Cohn of The New York Times, the data journalist, was interested in whether the data that I found could explain the Trump phenomenon. So I sent him my search data, and it was the single strongest correlation he could find with support for Trump in the Republican Primary. Trump’s map was very similar: Alabama, Louisiana, and Mississippi, and also Pennsylvania, Ohio, Michigan, and upstate New York.

Derek: And these were the states that he won by pretty wide margins.

“I sent him my search data, and it was the single strongest correlation he could find with support for Trump in the Republican Primary.”

Seth: And in the general election, he did better than a lot of people would’ve guessed, better than another Republican who didn’t say the same racially charged things and didn’t lead the birther movement for a while. You see in the data that there was a secret racism hurting Obama, and then this same secret racism carried Trump through the Republican Primary and helped in the general election.

Derek: Right. There’s this invisible force that you couldn’t necessarily see when you just ask people a question like, “Are you racist? Do you hold racist beliefs?” So one of the exciting things about your work is that it helps us to see the [nonsense] of so many stated preferences and surveys. It says you can’t just ask people what they think. You can’t just ask people who they are. They reveal it.

Seth: People can say one thing and mean something completely opposite. If you ask people in Mississippi or Alabama or Kentucky, “Are you gay?” very, very few men say that they’re gay. But gay porn searches are almost as high in these states as they are anywhere else.

Derek: This is clearly a negative implication of your research, right? That lots of voters are more biased and more racist than they might reveal in, say, a Gallup poll. Are there implications of your research that political scientists could interpret positively, that could provide a solution to a problem that we have in this country?

Seth: Yes. Because this data is so honest, you can find the problems that actually don’t exist. If African Americans, as I mentioned, think that many of their neighbors are secretly judging them, that’s correct. If Jewish Americans think that, that’s just not true in the nationwide data. My neighbors are not searching bad things about Jewish people, and if they really hated Jewish people, that would show up in the data.

Derek: So the data can reveal both the problems that we have and the problems that we thought we have but actually might not.

Seth: Right. [And] you can potentially change some of these negative attitudes. One thing I talk about is Islamophobia. Searches like “I hate Muslims” or “Muslims are evil” predict hate crimes against Muslims. They’re not just innocent searches. But with the Google data, we can actually break them down minute by minute and [see] what leads people to make more or fewer of these searches. I’ve analyzed a couple of Barack Obama’s speeches and seen that there are some things that inflame a mob, and some things that calm a mob down.

Derek: If I’m a speechwriter, what do I learn from an Obama speech? How do I calm people down?

Seth: After the San Bernardino attack in December 2015, Obama gave a speech trying to calm down anti-Muslim sentiment. It was nationally televised, and it was a beautiful speech, classic Obama. He said how it is the responsibility of all Americans to not give in to fear, to appeal to freedom. It is our responsibility not to judge people by their religion. You see in the search data how the searches changed for “I hate Muslims” during and after the speech. They didn’t drop, they didn’t stay the same. They actually went up. So basically, it seemed like everything Obama was doing was backfiring. But then he gave one line where he said that Muslim Americans are our friends and neighbors, they’re our sports heroes, and they’re the men and women who die for our country. And right after he gave this line, you see searches shooting up for Muslim athletes and Muslim soldiers, and people all over the internet are talking about, “Shaquille O’Neal’s a Muslim! I didn’t know that!”

speech

Compare the two strategies. One is a lecture, talking about responsibility, giving people information they already know. And the other is provoking their curiosity, giving them new information, changing how they think about a group that’s causing them so much animosity. So we suggest this, we publish this in The New York Times, and then two weeks later, Obama gave another speech in a Baltimore mosque. I think his speechwriters had read our article, because there was no talk of responsibility. It was all curiosity. It was all these facts about Muslim Americans, how they’ve built the skyscrapers of Chicago, and how Thomas Jefferson had a copy of the Quran. And after this speech, the searches for “I hate Muslims” went down.

Obviously this is [only] two speeches, so you don’t want to say that we’ve solved hatred in the United States. But it is suggestive that you can actually turn this into a science.

Derek: What’s really cool about the book is that it has these little action points where you see how to move people by using Google searches as a kind of real-time seismograph.

Seth: Yeah. Knowing the world’s problems is the first step towards solving the world’s problems.

Derek: Speaking of problems, let’s talk about relationships. One thing I thought was really interesting was the difference between what men search for online about their spouses versus what women search for about their spouses. What is this husband-wife differentiation here?

Seth: There are lots of interesting differences. One of them is that when men complain about their wives, they’re about three times more likely to complain that their wife is crazy than boring. When women complain about their husbands, they’re about three times more likely to complain their husband is boring than crazy. Women seem to want more excitement in their marriage, and men want less excitement, more stability in their marriage.

But the number one complaint that everybody has about their partner, whether it’s husband, wife, boyfriend, or girlfriend, is that the partner won’t have sex with me. That easily beats the second complaint, which is that the partner won’t text me back. One thing that I find really interesting: there are twice as many complaints that “boyfriend won’t have sex with me” than “girlfriend won’t have sex with me.”

Derek: Which surprised me when I read that.

Seth: But I don’t think we can go from that data to [the idea that] there are twice as many boyfriends withholding sex than girlfriends. It could be that because it’s more surprising [when men withhold sex,] women are more likely to turn to Google. If a girlfriend doesn’t want sex with a man, a man is just gonna tell his buddies and be like, “What should I do?” But maybe if a boyfriend doesn’t want sex with a woman, a woman’s embarrassed. She’s not gonna talk to her friends, [so] she turns to the internet. But I do think that this shows that young men avoid sex more than we realize.

“When men complain about their wives, they’re about three times more likely to complain that their wife is crazy than boring. When women complain about their husbands, they’re about three times more likely to complain their husband is boring than crazy.”

Derek: This gets at a larger question: to what extent is Google reality? It might not be the full picture of our feelings. It might over-index those feelings we specifically withhold from conversations, right? Sex is a subject that we are a little skittish talking about, which might be why there is so much about sex that we type into that little white box. Would you say that Google over-indexes things that we have to withhold?

Seth: I think it does, [though] I think Google’s closer to the truth than most sources.

But you do have to be careful turning from Google searches to an exact number. For example, I don’t think you can say from how much gay porn is searched in Mississippi exactly how many men are gay, because you don’t know how many people are making the same search. But you can definitely say that there are more gay men in Mississippi than surveys say.

Derek: You have a fascinating story about what Indian men search for.

Seth: In the country of India, the top search that starts, “My husband wants…” is, “My husband wants me to breastfeed him.” Which totally comes out of nowhere, and it’s way higher in India than in any other country. Porn that Indian men watch is dominated by breastfeeding. [Overall,] adult breastfeeding is a big theme.

I looked more into this recently, and if you look at the United States or United Kingdom or most every country, not surprisingly, searches for “How to breastfeed” are all like, “How to breastfeed my kid.” What you’d expect. And then in India, [searches] are equally split between, “How to breastfeed a kid” and “How to breastfeed a husband.”

Derek: That is amazing.

Seth: After I published this, someone interviewed people in India, doctors and Indian men and Indian women. Nobody said they knew anything about this, it’s not talked about. I Googled Indian breastfeeding, and my story came up, [but] nothing else.

It’s so wild, not just for this but for what this says about society. That there can be a widespread desire without reaching the level of public awareness. It’s not like one Indian man was breastfed by his wife and started telling all of his buddies, because nobody’s talking about it. So what caused it?

“We shouldn’t be so afraid of our weirdnesses, because our weirdnesses are actually far less weird than we think.”

Derek: It’s an incredible story, and it speaks to this point that Google data is best used not to make numerical estimates, but to tell us what questions to ask, right? You look at this data and you realize there might be this massive, unspoken fantasy that isn’t on any researcher’s radar.

I want to talk a little bit [about] medical concerns and self-diagnoses. You have a fascinating section on pancreatic cancer. Tell us a bit about that.

Seth: A study by researchers at Microsoft in Colombia is one of the most amazing studies I’ve ever seen. They studied Bing data, the same individuals over time, anonymous and de-identified. [They knew some of these individuals] got diagnosed with pancreatic cancer, [because] when you get diagnosed with pancreatic cancer, you probably search that in Bing. It’s a traumatic period in your life, so you look for support groups or whatever.

Then they went back and looked at these people’s searches in the months leading up to the diagnosis. What symptoms were they searching? [It turns out] their symptom patterns predicted they might have pancreatic cancer. They found these really subtle patterns, [like] indigestion by itself is not a risk factor for pancreatic cancer, but indigestion followed by abdominal pain is. These things are beyond what doctors know.

Derek: I was thinking that for Apple Watch users, that watch could send alerts. It might freak you out to a certain extent, but you could somehow raise concerns that say, “Hey, your search history of the last two weeks actually tracks several possible maladies. Let us know if you want us to alert your doctor.”

Seth: Or you could opt in. I would want to know if I had a 10% chance of [having] pancreatic cancer. The earlier you get diagnosed with pancreatic cancer, you can dramatically improve your odds of survival, so if I can double my odds of survival by being told, “You have a 10% chance of pancreatic cancer.” It would ruin my week, but I’d want to know that.

Derek: Absolutely. I think the undercurrent of [your] book is that we reserve lots of socially embarrassing questions and admissions for Google. [But] nothing is socially embarrassing when it’s talked about a lot. A somewhat inspirational element of the book is this idea that we shouldn’t be so afraid of our weirdnesses, because our weirdnesses are actually far less weird than we think. [Maybe] every month, someone should take a question that they reserve for Google and ask a friend.

Seth: I like that. We’re all kind of weird, and we should just embrace it.