Pushing your own understanding

A couple of days ago I finished Timur Kuran’s “Private truths, public lies”. I found it interesting but also very dense, with a very detailed explanation of a mathematical model that underpins most of the argumentation. I was under the illusion that I had understood most of it, or at least I had got the gist: people falsify their inner preferences to conform to a majority position, and this creates all sorts of distortions in the ideas that are discussed in society. Turns out there is much more to it. I gave the epub to Fable 5 (just before the ban; I had to continue with Opus 4.8 after it), so that it would quiz me about the book and, oh boy, I was wrong. I had understood a very shallow version of the arguments but I had missed some of the most central insights and mechanisms at play. I explicitly asked the LLMs to be strict and poke holes in my understanding, and in a long Socratic conversation I managed to have a few “aha!” moments.

Reading a book alone gives you about 10% or 20% of the gains. You get to spend some time thinking about the arguments, the anecdotes, the models: after that, you manage (with a bit of luck) to remember the main topic and a few of the examples.

A few years ago I discovered (yes, I know, who would have thought about it) that taking notes after reading the book or preparing some Anki cards gives you a much better understanding of the content and allows you to retain much more. Maybe the gains are about 50 or 60%.

This, discussing the book deeply with someone (something?) that “actively” tests whether you have got a grip on the models presented there, yields gains that are probably 80 to 90% of what you can extract from a book.

Of course, this slows down my reading rhythm a bit but I prefer quality over quantity, so I am certainly acquiring this new habit of having a deep discussion with the LLMs after reading something meaningful.