There are a lot of books in the world. How do you choose which ones to read?

There are at least two questions here:

  • What topics should I read about?
  • What books do I read within a topic?

For some people, the answer to the topic question is easy. They're obsessed with a given topic. They move onto the second problem.

For others, like myself, the first can be daunting. It's not that there aren't enough topics, it's that there are too many. If I look at my Goodreads want to read list right now, I see books on the following:

Each of these topics seem great and the books seemed good enough to justify saving. I have 176 more on my list. A friend just emailed me about Why Liberalism Failed and it turns out that book looks interesting as well. How to choose?

This post is about modeling my reading engine. It's about reading for insight, reading to uncover valuable and relevant information and skills. There are other reasons to read, like to relax or for social reasons. They can be related, but I'm largely going to ignore those reasons here. This may come across as utilitarian and boring, but that's the cost of being serious.

Let's step back and consider what kind of decision is at stake here. Decisions can vary on the following dimensions:

  • reversibility
  • stakes
  • prior uncertainty
  • post certainty

When decisions are irreversible, you'll want to ensure that you take the correct one. So you'll spend more time on the decision. If they are reversible and you learn you made the wrong decision, no problem, you can change. Choosing between job offers is the sort of thing you'll likely want to spend a decent amount of time on. The switching costs are usually high. If you want a better example, consider the decision of whether to have a kid or not.

If decisions are high stakes, deciding correctly matters more. Not so much if they're smaller potatoes. Spend more time on choosing a job, less time on what to watch this evening.

If decisions are made in a shroud of uncertainty, then the value of additional information will be higher. Prior uncertainty refers to the amount of uncertainty before a given decision. Often, "When it comes to perceiving reality, our minds are in a fog."

Post certainty refers to the amount and rate of relevant information flow after making a decision. That's a bit abstract, but we all know that after some decisions, you'll know right away whether you made the right one. Ordering at some restaurants is like this. After other decisions, such as longer term investments, you may never know. To be more precise, a decision will be higher in post certainty to the degree that, after you make the decision, you get relevant information and you get that information quickly.

With these in mind we can see that the decision to read on a particular topic is:

  • reversible
  • medium stakes
  • variable prior uncertainty
  • reasonable post certainty

The decision is reversible because you can stop reading at any point. Reading is somewhat important (it depends), but not as high stakes as many decisions we make. There can initially be a lot of prior uncertainty, ideally this goes down over a life. Occasionally, the result of a decision won't be evaluable for years, for example, when it's tied up with a risky research project. An example of this is deciding to become an expert in nanotech in the early 2000s. Yet if you know that you want to take a bet on high risk, world shaping technological developments then, by asking the right questions, you'll be able to reduce uncertainty. Given that you're studying nanotech, you can evaluate whether you're studying the right issues within that domain. You can do all of this even if the decision to study nanotech won't be evaluable for years.

So choosing what domains to hop into is reversible, medium stakes, has variable uncertainty, and has a reasonable amount of post certainty.

There are two key upshots here:

  • First, what you read is determined by your projects. If you're uncertain about your projects, that uncertainty will trickle down. When I was studying philosophy, learning software engineering, or heads down in Stoa I had less uncertainty about what to read. If I'm exploring different projects, there's more uncertainty.
  • The second, is that deciding what to read is a game you'll play over and over again. It's a reversible decision that you can get better at.

What you want to think of doing then, is designing a reading engine. You're not making one-off judgements about what to read for all time. Instead, you're setting up a system for selecting, consuming, and then utilizing information. Books play a small role in that ecosystem - they're not even the primary unit.

Building A Reading Engine

This engine can be modeled as a pipeline:

Created with Roam Research's nascent diagram tool.

Reading is the core activity. But it's interwoven with a number of other activities.

What I read is influenced by what I've collected into my reserves, projects, the questions I'm asking, and curiosity.

The reserves is a list of books that I found intriguing at one point or another. It isn't a list of books to read. That seems too prescriptive - I won't read every book in my reserve. Instead, it's a list of books that I may read, if I get excited about the relevant topic. The books are on deck, as it were.

Some of my favorite ways to populate my reserves is through interviews, blog posts, conversations, and Twitter. I usually find the best pieces at the end of bibliographies or during a simple search.

The most influential driver of this pipeline is projects. I like to think of my life as organized around a few projects or goals. Some of these are mundane, others grand. Whatever they are, they should guide what you read. If you're looking for insight, that's in the purpose of something or other. Examples of projects are:

  • Building better habits
  • Learning about formal methods in cultural evolution
  • Aiming to be more Stoic
  • Cooking more varied dishes
  • Running a reading group on power and influence in organizations

As noted above, uncertainty about your projects trickles down. If it seems like there are too many things to read, that may be a sign that you have too many projects.

If this is you, you can let your work inform whether they should get dropped or not. If you want to cull projects more aggressively, derisk and prioritize. Focus on questions like:

  • What can I learn that would make this project not matter to me?
  • What cheap tests would show that this project won't work?
  • Can I build an MVP of this project?
  • What specific questions about this topic do I want answered? What would answers unlock?

Projects are vague and not sufficient for narrowing down the solution set on their own. Instead, list questions that you want answers to. If reading is inquiry guided, you're more likely to find what you want and spend less time trawling through books that feel relevant, but aren't.

Questions can help you choose the right things to read. What you decide to spend your attention on is a reversible decision. You can stop reading at any moment. So, if, like me, you feel like there are too many topics, narrow it by answering questions like:

  • What's the chance that this topic is related to my project?
  • If I understand this completely, how will that be helpful in advancing my project?
  • How long will it take you to understand this topic? Is it worth the cost?

Think of reading as inquiry guided. Your questions are the fundamental units. Read in order to answer specific queries you have. Once you've answered a question, that should generate more questions and output. If you can't find the answer, put the book down.

As an example, I want to read the book on neural design, because I'm interested in the nature of emotion. Often I feel like I read neuroscientific explanations of behavior and they don't feel like explanations at all. On the surface, they appear to be very black box (x module is linked with this behavior). Plausibly, explaining how the brain works from an engineering perspective is more illuminating. So, if I read that book then I should come into with the questions:

  • Does this book give a new kind of explanation?
  • Is this explanation useful to understanding the nature of emotion?

I can likely figure this out pretty quickly. If the answer is no, I can stop reading. Returning to the framework above, reading has reasonably high post certainty. As you read, you can get a sense of whether you're reading the right thing by updating on the book’s quality and whether it has what you want in it.

As an aside answers to these questions are things that should end up in notes. And if the answers help me push the peanut forward, they'll influence other questions, an essay, or project.

Your project likely has some private or public output.

For me the private output is in the form of anki cards and notes. Anki is a fun and useful way to remember things. A lot of people are excited about note taking systems and software today. The most important part about notes is that they actually answer the questions that you have.

Public output can be anything from behavior change to essays, from conversations in person to email. Putting things out there will help you get feedback, speed up your projects, ask better questions, and provide value to others.

Occurrent Curiosity
Not all things should be so structured. Plausibly some amount of reading shouldn't fit that tightly into your system. You're more likely to discover something neglected if you read things for the sake of reading them on occasion.

Sometimes it's fun to read something because you're curious about it at that very moment. This is less systematic but I think that occurrent curiosity plays an important role in choosing what topics to read on. For me, it serves to explore books that may not be directly related to a project of mine, but could be relevant. Sometimes a particular question, book cover, or area catches your attention and it can be rewarding to follow up on it right away.

I expect that reading for the sake of reading should take up the minority of your time. It's driven by availability and recency biases. And if you do it too often, you won't be pushing your projects forward or answering questions you have.

However, occasionally you'll stumble on something that others have yet to understand and this can be quite great.

How to Choose

In some ways, this essay didn't answer the questions it set out to:

  • What topics should I read about?
  • What books should I read?

Instead, the claim is that these questions are the wrong ones. Uncertainty about what to read can be reduced by clarifying your projects, explicitly asking questions, and treating what to read as a reversible, low stakes decision that you'll get better at over time.

This framework is useful for ensuring that you're reading in line with your projects. With questions as the fundamental unit of your reading adventure, you can ensure that you're getting better answers and generating better questions.

Instead of asking, should I read this should I read this book on How To Think Like A Roman Emperor or this The Practicing Stoic? Spend a few minutes clarifying what questions you want answers to. Questions like:

  • Who was Marcus Aurelius?
  • How do psychotherapists apply Stoicism in their practice?

Suggest reading How To Think Like A Roman Emperor, while questions like:

  • What did the Stoics believe?
  • How do the main Stoic philosophers relate to each other?

Suggest reading The Practicing Stoic. If your questions are much broader like:

  • What even is Stoicism?
  • How can I apply Stoicism to my life?

Select a few books on the genre and briefly leaf through each of them. You're not choosing to commit your life to finishing any given book! Check out a clump, read a few, check in with your questions, return to reading, repeat. This is a process where you're getting feedback between your projects, questions, reading, and output.

Build a reading engine. Returns to systematicity are high.

Thanks Austin Wilson and Evan Gaensbauer for reading and commenting an earlier drafts.