Everybody Lies

The upshot of this book is not that big data is the holy grail. Rather, the recurring theme in all of Stephens-Davidowitz's interesting examples is just that most self-reporting is awful.

I'm still skeptical about the big data revolution–and this book doesn't really focus on implicit bias in analysis of large data sets–but the conventional research methods of social sciences are amusingly torn to pieces (much like advertising ROI was absolutely shredded in the digital age where measurement was no longer entirely by gut).

It was touch and go with this book. I was halfway into it and ready to put it aside, funny curiosities about what people search on Google or what Facebook data says about us simply doesn't cut if for me anymore. I think Yuval Noah Harari set the bar too high in Homo Deus, by covering most interesting findings big data has to offer in an elegant and concise manner.
However, the second half of the book covers a bit more of human behaviour, data misrepresentations and the extent to which big data can be applied. I found that to be unique and informative enough to do some research on my own.
Although Big Data is no longer considered to be the new kid on the block, it is still very important (Google and Facebook are definitely making a hefty profit out of it), but the sooner fields like psychology, sociology, anthropology are going to embrace it, the better for us as human beings.

Turns out the google searchbar is the only place where we don't lie to ourselves or others. That's where we reveal our strangest fears, hidden prejudices and kinkiest wishes. Data scientist can wrangle that data to learn how racist we truly are, how sexist we raise our children and how much sex we're truly having. The book contains a lot of examples demonstrating how big data and data analysis methods now enable us to see connections and correlations where previously we were just blind guessing. It helps to pick the winning racehorse, reveals that good students have promising futures no matter which school they attend, and shows that violent movies help lower the crime rates on opening weekends.

So, all in all lots of interesting anecdotes, a lot of them on the juicy side, mixed with occasionally slightly inappropriate jokes. The book could have used a little less of the author trying to insert himself. Also, some of his analogies were just wrong (no, the grandma is not big-data).

3.5

The big data equivalent of Freakonomics. Some interesting tidbits about conclusions reached through analyzing large data sets. Pretty good audiobook to listen to on the treadmill.