Andrej Karpathy's Reviews > Thinking, Fast and Slow

I could not bring myself to finish this book. The book is filled with shady experiments on undergraduates and psychology grad students and wild extrapolations of the associated results. I find it exceedingly difficult to take many of the conclusions seriously. I can't read into them. I can't trust them. I can't base my decisions on them and I resist incorporating them into my world view with anything more than 0.01 weight. In fact, several of the experiments that this book mentions were also found to be not reproducible by a recent meta-study on reproducibility in psychology studies.

Here's a characteristic example of me reading the book. The author says: "Consider the word EAT. Now fill in the blank in the following: SO_P. You were much more likely to fill in the blank with a U to make SOUP than with an A to make soap! How amazing. We call this phenomenon priming, system 1, something something". In fact, no, SOAP came to my mind immediately.

All I could think about when I read this book is my own experience of participating in a friend's psychology study once. He designed an experiment and asked me to do some things and answer some questions, but at some point it became extremely clear to me what the experiment was about, or how he hoped I would behave. I went along with it, but I couldn't believe that this would eventually become part of a paper. It was a joke. I'm afraid you can't go through a similar experience and take these studies seriously from then on.

All that being said I do find the broad strokes of the system1/system2 division proposed in this book to be interesting and appealing. A small few of the examples were fun to contemplate, and it was okay. 3/5, aborting reading.