Everybody Lies: Big Data, New Data, and What the Internet Can Tell Us About Who We Really Are

This book was fun, but it felt all too often that the conclusions it draws from the data it presented, didn't really follow from the data - or rather, it does to an extent, but with considerations.

This is a relevant passage from the book:

You might think that a terrorist attack that kills dozens or hundreds of people would automatically be followed by massive, widespread anxiety. Terrorism, by definition, is supposed to instill a sense of terror. I looked at Google searches reflecting anxiety. I tested how much these searches rose in a country the days, weeks, and months following every major European or American terrorist attack since 2004. So, on average, how much did anxiety-related searches rise? They didn't. At all.

The book relies heavily on Google search data and draws all sorts of conclusions from them, making the implicit assumption that Google searches follow people's feelings and opinions. It also assumes that people are more honest with Google than with surveys, for various reasons (which I would agree with).

However, I would think it's obvious that people's Google searches (and many other such datasets used in the book), often can't follow what people are thinking or feeling very closely. The reason is that Google is a way to research something, or to remember something. People don't Google all of their feelings, thoughts and worries, even though the book posits that they Google more of them than we tend to think. To illustrate simply, towards the above quote, after a terrorist attack, people might be scared to go to crowded places, they might feel anxious, they might fear for their loved ones safety. I'm not sure any of these necessarily lead to Google searches. People will probably Google the event, they will read about it in the paper, they will adjust their schedules accordingly to perhaps go to the place the event happened to leave flowers, or avoid it altogether out of fear. They might make an extra appointment with their therapist, or call their parents or their friends. They might even Google something (like the way to the place of the event). But lack of Google searches related to anxiety does not mean a lack of anxiety.

A classic (albeit simplistic) example is the amounts of Google search results for "gay sex" compared to those for "straight sex". "gay sex" returns about 30% more results than "straight sex" as of the time of this writing. Does that mean people have a lot more gay sex than straight sex? Or that people think a lot more about gay sex than straight sex? No. Partly, this can be explained by the fact that straight people probably don't qualify their searches with "straight". They largely search for "sex". Or, well, they search for something more specific, as Googling for "sex" probably isn't something people do when they think about sex.

Similarly, there was another passage conjecturing about suicide, which mentioned the rates at which people Google searches related to suicide. The chapter also talked about how actual suicide rates are far lower than Googling for suicide, which means many people consider it without following up. To which I immediately think of many ways the group of people considering suicide and that of people Googling suicide can diverge: some people just Google suicide or ways to commit suicide out of curiosity. Some, who actually consider suicide, will Google it a couple of times, and then find information on it which they can keep - and thus not have to Google for it again. In general, when people Google, they often find a community or a website which can solve their questions on something that interests them, and they bookmark or subscribe to that website. Or they keep the information and don't need to ask Google about it again.

It's a bit like measuring how many times a certain theatre is visited, using a dataset with all the occasions where people ask someone on the street for directions.

One can claim that people mostly don't ask for directions anymore (probably because they can Google their way to the theatre). Thus, my comparison is unfair, as people Google their thoughts a lot more than they ask for directions.

This highlights my opinion on the matter: do we really know to which extent people Google their thoughts? Is it the same per topic? One thing is for sure: there's a lot of phenomena that explain why real life conditions cannot be construed from Google searches, and until these can be measured, quantified or explained away, any conclusions drawn from such data is inaccurate in unspecified ways, and shouldn't be trusted.

All in all, read this book because it presents fun data. Don't expect much from the conclusions.

some people, for example, simply Google "is my husband gay", which is clearly something Google cannot answer for that specific husband, but simply reveals what this person worries about.