...and a follow up to my recent post “Recommended Readings.”

Introduction

Andy Matuschak’s 2019 essay “Why books don’t work” has been on my mind for a while now. I highly recommend you give it a read. The piece’s central claim is that books fail at their fundamental implicit task: conveying information to their reader. Let Matuschak speak for himself:

Picture some serious non-fiction tomes. The Selfish GeneThinking, Fast and SlowGuns, Germs, and Steel; etc. Have you ever had a book like this—one you’d read—come up in conversation, only to discover that you’d absorbed what amounts to a few sentences? I’ll be honest: it happens to me regularly…
I know I’m not alone here. When I share this observation with others—even others, like myself, who take learning seriously—it seems that everyone has had a similar experience…

Why? Because

books have no carefully-considered cognitive model at their foundation, but the medium does have an implicit model. And like lectures, that model is transmissionism. Sequences of words in sequences of lines in sequences of pages, the form of a book suggests people absorb knowledge by reading sentences.

But transmissionism is basically false. Matuschak concludes that

as a medium, books are surprisingly bad at conveying knowledge, and readers mostly don’t realize it.

And the answer, he thinks, is not to change how we read but rather to replace books with a medium more conducive to understanding.

We agree so far

I basically buy the central thrust of Matuschak’s argument. A typical nonfiction book contains probably many thousands of individual statements and propositions, the vast majority of which the vast majority of us will never remember. And I agree that changes to the medium can help. For example, Spencer Greenberg’s excellent post “How You Can Gain Self Control Without ‘Self-Control’” includes built-in, interactive flashcards to help readers absorb and remember its content.

Books do work, sort of

Matuschak acknowledges that books aren’t useless*:*

I’m not suggesting that all those hours were wasted. Many readers enjoyed reading those books. That’s wonderful! Certainly most readers absorbed something, however ineffable: points of view, ways of thinking, norms, inspiration, and so on. Indeed, for many books (and in particular most fiction), these effects are the point.

But I do think that he gives short shrift to what seems to me books’ hidden function: making an idea or a set of ideas take up an appreciable amount of room on our cognitive real estate. And I don’t just mean for more abstract “points of view, ways of thinking, norms, inspiration, and so on” but also for plain old information, both uncontroversial statements of fact as well as more contentious claims or arguments

Against the ‘Light Switch’ model of knowledge

Matuschak’s essay implies something like a ‘light switch’ model of knowledge, and this probably works pretty well for, say, state capitals and other “trivia”-style facts. You know the tenth digit of pi, or you do not. The switch is on, or it is off. For so much else, though, a statement’s salience is more important than whether we technically do or do not “know” it.

An example

I recently finished The Precipice by Toby Ord. His headline claim is that humanity faces greater than a 15% risk of “existential catastrophe” in the next hundred years. Before downloading the audiobook, I already “knew” this statistic. In Matuschak’s model, the flip for this proposition was turned to “on” in my brain, in the same way that I “know” that Annapolis is the capital of Maryland.

Further, I have either already forgotten or will soon forget at least 95% of the book’s ancillary facts and arguments. Right now, I do remember Ord’s description of a few terrifyingly close nuclear calls during the cold war, that he thinks unaligned artificial intelligence constitutes about 10 percentage points of the 16.6% risk, and a few of his proposals to improve international coordination among other things. If I were to enumerate every single fact and argument I remember, though, I have no doubt it would be far fewer than 5% of all those in the book.

But most of the book’s impact comes not from flipping any of these “fact switches” to ‘on’ - it comes from making that “one in six chance of extinction” much more salient to me than it was before. How does it do this? I’m not a psychologist, but it seems likely that 99% of the substance of the book is basically a trick to get me to mull over a cluster of ideas for a while.

Before, I knew that a very smart person whom I admire and respect thinks that humanity is basically playing Russian Roulette. Now, I really know that a very smart person whom I admire and respect thinks that humanity is basically playing Russian Roulette.

This is important! Lots of people “know” that climate change is a big deal, but not enough people really know that climate change is a big deal. In more precise language, the salience of climate change is probably relatively low even for those who are seriously concerned about it.

To most of us, I’m guessing, COVID is a much more salient crisis. Perhaps this is appropriate, but I doubt it. What might increase the salience of climate change? Spending ~10 hours reading or listening to David Wallace Wells explain why (as the first line of An Uninhabitable Earth reads) “It is worse, much worse, than you think.”

Another one

The very first “serious” non-fiction book I remember reading on my own is A Hope in the Unseen, probably 12 or 13 years ago. Despite its 384 pages and likely tens of thousands of individual propositions, my entire memory of the book is little more than the single sentence a smart poor black kid struggles to get into college, but is finally admitted to Brown.

Under the ‘light switch’ model of knowledge, it would have been much more efficient to simply memorize that italicized sentence. In fact, I “know” more information in the book after skimming the Wikipedia page just now than I remembered from reading the whole thing.

But that’s not the point. Because the book probably took me ~8 hours of reading spread over three weeks or so, my brain was forced to process a single, coherent narrative for quite a bit of time. The sentence “a smart poor black kid struggles to get into college, but is finally admitted to Brown” and a maybe a few other vague associations with this sentence have taken up a real amount of my cognitive real estate for the last 12 years. Not much—I don’t walk around thinking about Cedrick, the protagonist, all day—but more than could have been achieved by reading the Wikipedia article no matter how recently.

Books are clandestine salience-building devices

My hypothesis is that—in agreement with Matuschak—the thousands upon thousands of bits of information conveyed in a typical nonfiction book don’t effectively transmit themselves to a typical reader, but they do force the reader to marinate in the book’s main few ideas for enough time for something important to happen.

Another lens

To put words in someone else’s mouth, I think Matuschak would say that, for the purpose of conveying information, it would be much more efficient to read a very short summary than to read an entire book. After all, we never remember more than the summary’s content, and generally remember much less.

I think this is incorrect.

After a few weeks, months, or years, a person who read the book won’t remember more than the short summary’s content, but the person who read the short summary won’t remember anything at all. Or, if the latter does remember something, the former will remember far more.

On one level, this is completely trivial. People remember more stuff if they learn more stuff, and a book has more stuff inside than a summary. Duh.

But the key point to remember is that (to a rough approximation), reading a book doesn’t convey any information not included in the summary. All the thousands of extra sentences and anecdotes and facts aren’t there to be remembered. They’re there to trick the reader into spending enough time mentally interacting with a few key ideas.

So, if you’re an author who wants to convey an actual set of, say, three or four main ideas, you could write a couple dozen articles or a couple hundred tweets and hope that a reader decides to read every single one. Or, you could write one book, filled with tens of thousands of quasi-redundant statements that will rapidly be forgotten, that you can reasonably expect a good proportion of initial readers to fully consume.

Mechanisms

At a psychological or neuroscientific level, I don’t really know what’s going on, but I can speculate. We know that availability bias disposes us to treat the ease of recall of a memory as a proxy for its importance, and the quantity of recalled events as a proxy for actual frequency. That’s why it’s surprising that about as many Americans die from asthma in a single year than have been killed in terror attacks since 1970 (more than 75% on 9/11).

Worth noting how similar the NYT and Guardian coverage is

Even for those in the intelligentsia who “know” that heart disease and vehicle accidents kill orders of magnitude more people than do terrorists, or that nearly two thirds of U.S. gun deaths are suicides and only 0.2% come from “mass shootings,” it is difficult to shed availability bias’s distorting influence. Mass shootings don’t intuitively seem like 0.2% of the problem. This isn’t to say that we should care about everything in exact proportion to its frequency or “objective” importance, but the two should probably be more closely linked than they are.

Back to books

Anyway, I think that books are basically mechanisms to leverage this availability heuristic. Even after reading that “mass shootings” (not sure how these are defined) constitute a tiny proportion of gun deaths, their disproportionate media coverage (perhaps for good reason - that’s not the point of this post) maintains their outsized share of our cognitive real estate.

The sheer length of time required to read a book is analogous to regularly watching violent crime reporting in the nightly news, albeit hopefully to a better end. You’ll rapidly forget 99% of the book’s statements just as you’ll rapidly forget 99% of the nuances of every news story. All those forgotten propositions are doing something important, though, for better or for worse: keeping you engaged with a few central ideas long enough for them to make an impact.

New to LessWrong?

New Comment
16 comments, sorted by Click to highlight new comments since: Today at 12:52 PM

Another way to think about books is as training data for a deep learning network. When you feed a thousand photos of airplanes into a neural network your goal isn't to get the network to remember the specific airplanes in your training dataset. If you did you'd probably have made the mistake of overtraining it. Rather, your goal is to teach the network to recognize the abstract concept of an airplane. Books could be similar. Their purpose could be to feed many variations of a single concept through a human brain so it can absorb the pattern.

Super interesting and likely worth developing into a longer post if you're so inclined. Really like this analogy.

The premise that a book can be summarized into a few key take-aways seems reasonable - non-fiction authors are required to summarize in that way as part of the contracting process, since it demonstrates the book has a clear set of arguments or aims. However, a primary function of the other sentences is to offer evidence that supports the central propositions. One could summate The Selfish Gene in a few sentences, but why should anyone be persuaded by those take-away claims without any evidence? We don't need to take away the bulk of the evidence in detail (the majority of sentences), but we do need the evidence in order to have faith in the key ideas.

A point that's been made to me before about books and that supports your thesis that they serve to increase salience of ideas: books are worth reading because they cause you to live with an idea long enough to integrate it into your mind.

I've certainly noticed this facet of books for myself. There's plenty of books I read that I could now hardly recall any specific arguments from and that I could summarize their "take-aways" in a few paragraphs, but that nonetheless I got a lot out of reading because for the weeks or months when I was working through it the book forced my attention continually back on to a particular topic, which then also carried to thinking about that topic at other times.

I carry this same sort of idea with me into my choice of where to place my attention beyond books with a question like: how will placing my attention on this bit of media affect what I pay attention to, and is it pointing my attention towards something I think is worth thinking about?

I agree that books both transmit knowledge (à la light switch) and help that knowledge to take up space in your salience landscape. I don't think it's fair to argue that Matuschak is implying an only-light-switch model. When discussing metacognition, they focus on understanding rather than knowledge. Some of their metacognitive strategies that enable understanding are about making connections between pieces of knowledge (light switches, in your model).

To put words in someone else’s mouth, I think Matuschak would say that, for the purpose of conveying information, it would be much more efficient to read a very short summary than to read an entire book.

Maybe. But what Matuschak does say is that information/knowledge would be better conveyed and understanding made simpler if ideas of knowledge conveyance and understanding were baked into the medium. Matuschak doesn't suggest that a very short summary is that medium.

Matuschak asks:

How might we design mediums in which “readers” naturally form rich associations between the ideas being presented? How might we design mediums which “readers” naturally engage creatively with the material? How might we design mediums in which “readers” naturally contend with competing interpretations? If we pile together enough of these questions we’re left with: how might we design mediums in which “reading” is the same as “understanding”?

That's not a light switch.

Yes, I was incorrect about Matuschak's position. He commented on reddit here:

"I think Matuschak would say that, for the purpose of conveying information, it would be much more efficient to read a very short summary than to read an entire book."

FWIW, I wouldn't say that! Actually, my research for the last couple years has been predicated on the value of embedding focused learning interactions (i.e spaced repetition prompts) into extended narrative. The underlying theory isn't (wasn't!) salience-based, but basically I believe that strong understanding is produced with a rich network of connections and a meaningful emotional connection, both of which are promoted by narrative (but usually not by a very short summary).

Thanks for the link! :)

I really like your way of thinking about why books are useful!

This reminds me of another argument for why books are useful which came up in this 80,000 Hours podcast episode with Julia Galef. 

Julia Galef: [...] You know, the thing that I think books do really well is provide a nice container for a thesis or ideas, such that it’s easy to spread and talk about. And they do this better than blog posts, for the most part. I’ve heard people sometimes say, “Most books should be blog posts,” or “Most books should be articles,” or something like that, and I sympathize with that view.

Another way of phrasing this: when two people have read the same book, even if they don't remember the details, they can reference the book as a "pointer" and make deeper arguments (held up by their intuitions about the book, ingrained because they spent so much time engaging with its entirety) than they would have been able to make if they had only read summaries. 

Yes this is an excellent point; books increase the fidelity of idea transmission because they place something like a bound on how much an idea can be misinterpreted, since one can always appeal to the author's own words (much more than a blog post or Tweet).

My essay How to Read a Book for Understanding was 90% motivated by annoyance at Andy's misunderstanding of how books are read differently by different types of readers. I never reformatted this for Less Wrong, but probably should. https://www.reddit.com/r/slatestarcodex/comments/d7bvcp/how_to_read_a_book_for_understanding/

I definitely agree though with your point about salience. I think it is an important though inadequate defense of books. Salience can be achieved in many more sophisticated ways than reading, even some YouTube videos create salience on surprisingly complex topics. Books offer more than just this as a medium.

Great post and thanks for linking to it! Seems like books' function and utility has gotten more attention than I would have expected. 

What blog posts are for: a response to "What books are for: a response to 'Why books don't work.'"

I read this blog post carefully yet absorbed only a small fraction of the total details it contains. You're only communicating one key idea here. For greater learning efficiency, you may as well replace this post with a one-sentence summary: "Anyway, I think that books are basically mechanisms to leverage this availability heuristic."

But then readers would have to repeat this sentence for as long as it takes to read the blog post to get the same effect. Not quite as fun.

Things that come to mind:

  • If we consider a textbook, then obviously the factual contents are usually way more than what could be summarized in one sentence.  "Twenty or so mathematical concepts, plus 1-5 ways to manipulate each concept and link it with the others", except that's not the factual content, it's merely a description of the factual content.
  • But also with a textbook, ideally, as you the reader go through it, you work with it: in a math textbook you work through examples or try to prove things yourself, in a science textbook you ask questions like "Is that really true?" or "How did they know that?" or "Does that mean someone could make technology exploit that effect?  Are they already doing this?" or other stuff.  This will (a) solidify the knowledge in your mind and (b) give you practice at thinking / investigating the subject matter.
  • Even if some nonfictional book is not a textbook and consists entirely of a thesis sentence and a collection of evidence and arguments to prove the thesis, you the reader can work with it: take each thing and ask how it might be wrong, if the evidence admits different interpretations, if there's a hole in the logic.  You can come away from it with (a) good practice at interrogating claims, (b) either familiarity with a good example of how to do investigation of that subject, or knowledge that the author was deficient and yet managed to get their book published; and (c) either a solid understanding of the important things connected to the central claim, or a list of holes the author didn't fill and that you might follow up on.

I think this ties in a lot to Matuschak's idea that it would be good to have an "authored time dimension" to books, sort of like how meditation apps spread out the experience of learning about mindfullness over several month. This makes learning about meditation from apps much more salient than reading about it from a book that you finish in the space of a few days. I find the idea that books should have an authored time dimension intriguing, and I think Matuschak's thoughts on the topic indicates that you overstate the case that he does not understand the importance of letting ideas "take up an appreciable amount of room on our cognitive real estate".

Books are best seen as a way to keep knowledge than a way to share it. Well, that's what they were designed for (that's what writing is for in fact).