[I explore how the nature of the essay medium presents potential challenges for rationality as a field.

First off, essays often aren’t updated and exploring new ideas is often more rewarding than revisiting old content. This means that there is not much of an incentive to create accessible textbook-like introductions to rationality.

Secondly, despite having ported over to a digital medium, essays are still largely the same. I think there’s value in exploring different norms and features in our writing.

Thirdly, I point out how, even if we figure out how to create very good online content, the way that the medium shapes our expectations might still not be good enough. Often, structure seems to be an important component, and I also don’t think we’ve pushed in this direction enough either.]

Rationality is concerned with a good number of questions. The two that I'm thinking about right now are:

  1. Why do some attempts at behavioral modification succeed and others fail, both between individuals and within an individual?
  2. If we had an answer to 1, what does it look like to actually make that knowledge usable, bridging the gap from just a mere description?

I think both of these queries connect to the question of how the nature of the medium affects our actual experience when trying to engage with rationality concepts. From a design and pedagogical standpoint, I think there changes we could make to potentially improve the ways that ideas advance and cash out into practicality.

Where Are All The Good Rationality Textbooks?

Let's start with something more personal.

I hate revisiting old essays.

I'm happy linking back to them, sure, but the idea of editing or rewriting them feels abhorrent. As a kid, I hated checking my work when doing math. In both cases, there's a feeling of "I just put in all this effort to get this thing done. Now, you want me to...do it all over again!?"

There's a sense of, not just our expectations being violated, but of all of our past work being essentially for naught.

For example, it's not exactly the marathon runner who is told they have 3 more miles to run when they thought they were already finished, but the marathon runner who is told they ran the race backwards and needs to start over from the actual starting line.

This sense of "not wanting to repeat things" is what I think forms a core part of why some rationality skills fail to stick. Once Murphyjitsu, Trigger Action Planning, Internal Double Crux, or any other technique of your choice is no longer shiny, it feels less tasty to use. This is the idea behind Fading Novelty, where we can become acclimated (and thus less driven) to pursue certain mental skills.

I think this sort of aversion is problematic when it comes to providing good explanations of concepts because it discourages an evolving revision process. It's not common for people to update most essays with the intention to increase how well it conveys its ideas—after its initial publication, that is. There's a manner in which most essays date themselves, acting as snapshots in time rather than an adaptive/improving/evolving set of ideas.

(Textbook authors are an impressive exception.)

This also lead to a weird first-mover situation I've noticed in the LW community where we have basically one canonical explanation of a concept, for example the “Politics is the Mind-Killer” that everyone links to... except that the original explanatory essay doesn’t actually go into that much detail.

As a concept, “Politics is the mind-killer” derives much of its meaning from additional essays which further develop the term and its implications. Someone in the community who’s been following the development of these ideas will have no problems following the established usage.

But for someone new to all of this? The original essay hasn’t aged to reflect the ways the concept has developed.

In other fields, we have introductory textbooks which are written specifically for the purpose of distilling concepts to newcomers. There's a reason we leave the original writings from Leibniz and Newton for historians and give intro calculus students something written later. It's rarely ever the case that writings which first introduce a concept will also end up being the best introduction to said concept.

Alas, writing improved expositions often feels less exciting than the novelty offered by improving existing ideas. Socially, it feels like there’s less of an incentive here for people to undertake this task. In other fields, while I think it’s also lower status to write introductory texts for other fields, but they do get written.

There’s something about the nature of ownership and ideas, where it feels a little odd to endorse someone else’s concept entirely; instead, I feel like there’s often a compulsion to justify how your concept is secretly a little different. Valuing ideas because you came up with them means that you now have additional reasons to hold onto them, outside of their merit alone. It can also cause us to misinterpret contrary evidence as a personal attack. (“You’re saying I’m <insert bad thing> for coming up with this idea!?” vs “You’re saying that the reasons for my holding this idea rest on faulty premises”.)

For rationality in particular, I think that there is no good standard set of essays which both:

1. Offers an accessible introduction to at least parts of rationality

2. Is periodically updated with better explanations

This leaves me wanting. My own attempt started off with my efforts to try and provide some new introductions to old concepts. But by the time the project was wrapping up, though, I also found myself cheating; to fill in later chapters, I copy-and-pasted in old essays with only minor edits for readability.

I’d like to find a way to incentivize the creation of better introductory materials, and I’m disappointed that there’s currently not as much revisiting / cross-dialogue between existing essays. It feels like lots of people are writing about things which, though they all fall under roughly under the same banner, aren’t building directly off one another.

Exploring the Essay

In a more general sense, I’m disappointed that more exploratory formats and norms haven’t been tried out with regards to essays.

One immediate area is that of leveraging the differences presented by putting essays on an electronic medium—there are a lot more opportunities afforded to us in terms of interactivity (e.g. user voting), formatting (e.g. highlighting examples), and metadata (e.g. word count).

But let’s leave that aside for the moment.

For the most part, I think most of the essays we’ve seen have been digital ports of the written word. Pictures, exercises, videos, and other types of content are scarce. It feels like, when communicating, we put in the least amount of effort required to conform to existing norms regarding essays. For example, we will use proper grammar/spelling in an essay, but not everyone will include a short summary of their essay before the main body.

I think that this has more to do with what other essays look like, rather than people deeming that [thing they’re not doing] isn’t very useful in aiding in communicating their ideas. When writing essays, and we are willing to as as far as required to comply with them, but anything more feels like “work” (whereas everything before it was just “necessary”.)

(If including a short summary in front of an essay feels redundant or unnecessary to you (EX: “Why read the summary if the essay is just below it?”), here’s a quick flip: Imagine a world where no one regularly capitalized things but it was instead customary to include short summaries before all your writing. If you suggested that we capitalize the first letter of every sentence, how would you respond to criticisms that capitalization was unnecessary because periods already act as sufficient delimiters?)

I think we can do a better job of trying out different norms for writing. For an example of a norm that did catch on and provide value, I point to how we started using the “Epistemic Status” tag in front of our essays. Since then, reading essays tagged with this piece of metadata has helped me find the right frame of mind to engage with said essay’s ideas.

I suspect there are other low-hanging fruits similar to this, like also including “Epistemic Effort”. Here are some potential ideas:

  1. Author annotations for the essay, in a style similar to Genius. (“If we can’t get the author to revise their work, maybe it’d feel less aversive for them to comment on what they now think / how things have changed?”)
  2. Crowdsourced summaries for essays. (“Can we leverage the comprehension of people who really ‘got it’ to crystallize what ‘it’ is for others?”)
  3. Guided discussion questions. (“Are there premises that are likely to be contentious, or are there areas the essays leaves unresolved?”)

While the typical essay format might be one of the best ways to describe something, I’m primarily concerned with the actual effect of an essay. I think this warrants trying to find improvements when, as I’ll try to show in the next section, giving someone a good description of X is not at all similar to teaching them to actually do X.

The Medium Really Matters:

Say we write up a model of how habits work:

“A habit is a stimulus-response pattern triggered by a predictable sensory cue, paired with an anticipated reward.”

Cool. Someone reads the above, and they now have an understanding of some of the moving pieces involved in making habits happen. Maybe they’re still unsure how to do it. How can we turn the model into something more concrete?

Say we break it down into steps:

  1. Identify an Action you want to do.
  2. Find a concrete sensory Trigger for the situation where you want the action to happen.
  3. Describe the Action you’d like to perform, in detail. Be specific about the action you’d like to.
  4. Put the Trigger and Action in a “When [Trigger], then [Action]” loop.
  5. Write the TAP down somewhere you can find it again.
  6. Physically or mentally (less good, of course) rehearse the TAP at least 5 times.

Great. After reading the above, someone now has a very clear picture of what the actions they need to take if they want to form a habit. But what might that actually look like, in practice?

Say we finally go into the details and give examples for each one of the steps:

  1. Go jogging more.
  2. The feeling of the coarse rope that opens my curtains in the morning.
  3. Pick up my jogging shoes and walk outside the door to begin jogging.
  4. When [I feel the rough coarseness of my curtain rope,] then [I’ll go grab my jogging shoes and open the front door].
  5. Having a digital or physical “TAPs List” document can make it easy to review which habits you’re currently training.
  6. Actually take the several minutes to do some visualizations—going over it in your mind helps you recognize the Triggers when they show up in the real world.

Fantastic. Now the reader has a very clear idea of what’s going on, what steps to take, and even has a working example they can tweak for their own purposes. We’ve given them a full set of the tools to get started making their own habits.

But, the question remains, will they actually do it?

I think the answer is still “No”. There’s something fundamentally off here, and I think it has to do with the sort of state people are in when they read this stuff.

I think people should be approaching this stuff the same way one might work through a math textbook—verifying the reasoning, doing the exercises, and trying to rederive the concepts themselves. At least for me, though, I think about LessWrong as a place where I go to unwind / read some fun essays; it’s not always a place where I’m ready to put in effort and direct all of my active attention.

There are ways that we can bridge part of this gap. Most of the tactics I have in mind involve going meta and/or using emotionally charged language to break the fourth wall and speak more directly to the reader. Essays on LessWrong often do this with some variant of “No, actually go and spend 5 minutes thinking about this!”

But even if we leverage all of the manipulative power of text, I don’t think that’s enough. When it comes to pedagogy, I think you can think of the multiple ways to learn things in terms of a gradient of increasing structure:

  1. Self-studying a textbook, which is completely self-reliant.
  2. MOOCs which provide some additional guidelines.
  3. Enrolling in an actual course, which has a fixed lecture time and strict deadlines.

As you bring in more structure, you start being able to take advantage of the additional social commitment mechanisms afforded by the classroom environment. MOOCs have assignments and often some sort of online support. Classrooms, of course, have the full social aspect of classmates, strict deadlines, and benefit largely from the social expectation that you School Is Something You Should Do As Your Student Duty.

And while we can levy a variety of criticisms against the typical school paradigm, I think it is at least somewhat effective at actually getting people to do things. Begrudgingly perhaps, but it does in fact instill compulsions to get the work done. There are, I claim, a lot of small crucial factors which aid in getting people to do stuff, which we abandon when we leave the classroom setting.

Under this sort of framework, if we look at rationality, we’re still at the point where basically the only way someone can learn this stuff is by reading a scattered multitude of essays (not even a textbook!) without much structure or a social aspect. If this is the default, I don’t think we should be surprised when people don’t seem to be able to easily put rationality into practice.

This is my take on why CFAR workshops stand a significantly better chance at teaching the “rationality thing” than any write-up. The written sentence “Now set a 5 minute timer to come up with reasons that you don’t want to X” is clearly going to have different effects than the setting where you’re in a room with other people and you’ve actually carved out the time to sit down and actually write down your reasons.

The choice of the manner in which information is transmitted has cascading effects, from the initial framing to the reasons that allow people to stick with continuing their efforts.


New Comment
7 comments, sorted by Click to highlight new comments since: Today at 11:45 PM

Edit culture could improve is a bit. As stackexchange allows suggestions to be submitted to the answers. We could do that. Implement "rewording suggestion" comments.

I'd contribute to a textbook. It's a hard project. At this point after writing 150k words I am finally hitting good quality. That would be bigger and harder than anything I've written before.

I'd also consider rewriting old content.

You also appear to have captured something odd. If people were able to solve their own problems, we would not be here talking about how to help people solve their own problems. The nature of the preface is that automagically expecting things to work does not work. Realising this, we need to develop means to break through that problem. That is already in the nature of the original problem we face.

Part of my thoughts on peer review is specifically to move to a world where a) it's considered a normal part of the writing process to rewrite it at least once, b) it's more expected for old, out of date essays to get revamped.

The lyrics platform genius.com used to want to branch out into education and its annotation function still seems like a valuable feature I would like to see on LW one day. This might ease constant discussion and development of otherwise static content by facilitating commenting, discussions and editing processes in very specified ways.

I suspect that a large part of the issue is that we don't have an organisation dedicated to spreading rationality in a wide-scale manner. Everyone in the rationality community tends to be rather busy, so I suspect that the only way to ensure that such tasks occur would be to actually pay people to complete these tasks.

Very interesting read. It looks to me like the post identifies 3 main problems with the essay medium:

1) The knowledge is dispersed and not indexed.

2) Knowledge becomes obsolete relatively quickly as new insights/better explanations are generated relatively frequently.

3) The context of communication around digital essays is not adequate for high-effort learning mode.

Modulo the motivating problem you mention, the digital textbook medium you propose seems to solve all three, assuming new, revised "editions" are released every now and then. Do you have any thoughts about how it compares with a Wiki medium?

I have recently started trying to organise my writings and I am also unhappy about the limitations of the essay medium. 3) is not an issue in this case, but I am quite bothered by 1) and 2). I also worry about:

4) The essay medium has a linear structure; expositive text generally does not.

It seems that the structure of expositive text is more similar to a tree-like graph: often there is an explanation about a particular idea X, and then the flow of the text "branches" into different topics to be discussed after introducing X (though paths sometimes rejoin). Perhaps this is not a big issue, but right now I feel like I'd prefer a medium that reflects this structure.

5) Explaining ideas in detail makes it harder to follow the big picture.

On re-readings of the material, sometimes I wish I could "compress" explanations I already understand to focus on the higher picture and relations between ideas. Skimming seems the thing to do, but it requires some sort of indexing, and right now I feel like digital media should be able to offer more powerful ways of doing this.

Posting essays on a wiki solves all the problems!

[-][anonymous]5y 2
Modulo the motivating problem you mention, the digital textbook medium you propose seems to solve all three, assuming new, revised "editions" are released every now and then. Do you have any thoughts about how it compares with a Wiki medium?

Hm. I think that Wikis might be easier to maintain, perhaps? There seem to be a lot of additional considerations when you're writing to teach and not just to explain / inform.

5) Explaining ideas in detail makes it harder to follow the big picture.
On re-readings of the material, sometimes I wish I could "compress" explanations I already understand to focus on the higher picture and relations between ideas. Skimming seems the thing to do, but it requires some sort of indexing, and right now I feel like digital media should be able to offer more powerful ways of doing this.

This matches my experience. I'd like some way of "collapsing" certain paragraphs, or labeling them as "basically just about X" to help with overall comprehension.

New to LessWrong?