The whole premise of the book is to discuss the disadvantages of “the self”. It defines “the self” as “the mental apparatus that allows people to think consciously about themselves” and attributes these abilities to it:

In the same section it also seems to equate the self with System 2 in general - “One important difference between automatic and controlled processes involves the fact that controlled processes require a self, whereas automatic processes do not. To make deliberate decisions to control their natural reactions, people must be able to think consciously about themselves and the implications of their behavior.”

So why did the author lump simulation, planning, executive control, introspection, empathy, identity and status into a single concept? I think the problem probably came right at the start - “I believe that the defining difference between human beings and other animals involves the nature of the human self.” The desire to find a single defining difference between two extremely complex information processing systems seems misguided, at best.

I notice computational approaches to cognition fall into this sort of trap less often because it distances us from our built-in mind-simulating machinery. If we think about the list of abilities above from the point of view of an information processing system, it seems far less clear that the word ‘self’ refers to the same representations or processes in each one.

For example, we can imagine a mind that can make accurate theories/predictions about the knowledge and physical abilities of itself and others, but not about motives/values/personality. Such a mind would be able to make complex plans to deceive it’s peers (reasoning about knowledge), but would be constantly surprised by it’s own behavior (reasoning about motives). Does this mind have “the mental apparatus that allows people to think consciously about themselves”. It’s too fluffy a concept to decide.

(If this mind seems implausible, bear in mind that there is a fairly compelling argument that most human introspection is actually posthoc story-telling, and that we have very little access to our actual thought processes. Every time you make plans contingent on getting up early in the morning and then find yourself sleeping in, you have failed to predict your own behavior.)

The book then rambles on through the disadvantages of ‘the self’, such as self-serving biases, self-chatter, interference with practiced skills, many of which seem to have little connection even with the authors definition of self. It’s a morass of linguistic confusion, held together only by the fact that the language it’s written in came long before any kind of understanding of cognition and so happens to use the same word for many different capabilities and experiences. They intuitively feel connected, but it falls apart as soon as one tries to imagine the underlying machinery.

Not every book has to be rigorous, but this book doesn’t have anything else to offer. No interesting thesis. No actionable advice. Nothing to take away.