Essay-Writing Checklist
Specify
Explain each idea in simple words, no matter how obvious you think it is.
Use each idea 25 times to make sure you actually understand it, especially when you don’t want to.
Write concretely, like the singer Gaana Baala. Don’t just say “going up a fragile tree”, say “going up a drumsticks tree that will break if you climb it”. People in oral cultures seem to think too concretely; they apparently have trouble thinking in terms of abstract concepts. I fear that people who read a lot may have the opposite problem. We have our heads way up in the clouds, when we should be down on earth examining concrete specimens. PG is on to something here with all his everyday analogies. You never feel he’s writing for a niche audience - anybody can understand what he’s saying.
Have tons of concrete examples backing up each point. If you want to say, for example, that the scientific method is better than other techniques, show how it is better. Give real-world examples of it kicking ass.
Treat your essay like a hypothesis, not a position you have to defend. Don’t try to convince the reader about how epic awesome these ideas are. If those ideas really are powerful, it will show in the problems they can help you solve. If it doesn’t show, they’re probably duds - why waste time trying to oversell them?
Start from the evidence. This will jog your memory because you will be looking for concrete instances (particular books or blog posts) instead of abstract models (“locality of causality”). Plus, you can use those sources to verify your ideas. Then, once you’ve got all the raw material out there, you can shape it into a beautiful model that is probably better than the vague one you have in your head.
Surprise
What question am I trying to answer? More importantly, what is surprising about my answer? Maybe make sure every paragraph or two has some surprising twist, a la PG. This is not just for writing style. If you think the conventional hypothesis is good enough, why are you writing an essay? Show exactly how your hypothesis predicts differently.
Fail Faster! Remember, it’s ok to be wrong. In fact, it’s great! It means you’re getting more Predictive Power. You want to be wrong as soon as possible. So, stick your neck out. Make narrow, bold predictions.
Organize
Have a clear flow of ideas from start to finish. Condense the information so that it is easy to grasp. Turn it into a Gem. Have a few small chunks to represent complex ideas (“Make something users want”).
Compress each section and make that the lead or the ending. Aim for four high-level chunks.
When in doubt, summarize your essay so far. List all the surprises you’ve found so far. Then shape the essay till you have a clear logical chain through those surprises.
When I write in my journal, I summarize after a few pages of notes. This forces me to clean things up and get even more ideas. I don’t do that for essays. I just add the first draft of a section and leave it there, making the essay harder to read for me the next time around. So, summarize and rewrite each section before you move on from it.
Motivate
All this surprising stuff is fine. But why does it matter? What decision will I change because of the questions I’ve answered?
Style
Clear and simple writing: Read it out loud.
Read sentences out loud to see what parts are awkward and contrived. Maybe use the diction and style shell utilities.
I shouldn’t be able to hide poor ideas behind decorative elements. No lists, no bold words (sigh), no All-Capitalized Awesome Phrases, and no italicized words. Let your idea speak for itself.
I think a lot of PG’s brevity and punch comes from his use of anaphora. He writes English the way programmers write code - by removing all unnecessary duplication.
Quick example:
You probably need about the amount you need to go running. I’m often reluctant to go running, but once I do, I enjoy it.
…
Do something hard enough to stretch you, but only just, especially at first. If you’re deciding between two projects, choose whichever seems most fun. If one blows up in your face, start another. Repeat till, like an internal combustion engine, the process becomes self-sustaining, and each project generates the next one. (This could take years.)
He doesn’t say “but once I go running again” or “but once I try to get out”. He just says “but once I do”, which seems so much smaller and sweeter than the previous two. I suppose it’s because small chunks fit easily within our working memory. Similarly for “hard enough to stretch, but only just” - he leaves out the pointless “but only just hard enough to stretch you” and ends on a strong note. Again with “If one blows up in your face, start another.” Look at how short that second branch is: “start another” - that’s all. Not even “start another one”. Just “start another”.
I don’t fully understand this and I’m sure people have already analyzed this before. Anyway, that’s the level I need to aspire to.
Technical Details
URLs are case-sensitive - didn’t know that! A link with caps works but not the lowercase version: ./Spaced-Repetition-My-Principles.html vs ./spaced-repetition-my-principles.html.
I’ll need to make all my URLs lowercase now.
Pandoc and MathJax
MathML apparently works in HTML (in Firefox and Chrome) without my having to include the MathJax script.
Key idea: Look at the HTML output of your markdown file.
Pandoc transforms my AsciiMath stuff to poor MathML. I think that’s because it treats the stuff as TEX (User’s Guide). For now, just use TEX commands within dollar signs for your inline needs (Wikibook). For the rest, get and insert their MathML version manually.
Feed
For now, I just want to update the feed whenever I publish a new full essay. I don’t care about updating the feed for smaller changes. I don’t think there’s much value in that information for anyone.
Given that I only plan to publish a new essay once or twice a week, I don’t think I need an automatic feedmaker. I can update the feed.xml file manually.
Also, I’ll go with Atom for now, since it seems to be the way of the future. However, I’ll just go with the url as the id (Gwern seems to be doing that too, and he’s generating the feed automatically, so I guess I’m fine).
To get time for the “updated” element: use date +%Y-%m-%dT%TZ
Error 404 pages on Google Webmaster
Q: Most of my 404s are for bizarro URLs that never existed on my site. What’s up with that? Where did they come from?
A: If Google finds a link somewhere on the web that points to a URL on your domain, it may try to crawl that link, whether any content actually exists there or not; and when it does, your server should return a 404 if there’s nothing there to find. These links could be caused by someone making a typo when linking to you, some type of misconfiguration (if the links are automatically generated, e.g. by a CMS), or by Google’s increased efforts to recognize and crawl links embedded in JavaScript or other embedded content; or they may be part of a quick check from our side to see how your server handles unknown URLs, to name just a few. If you see 404s reported in Webmaster Tools for URLs that don’t exist on your site, you can safely ignore them. We don’t know which URLs are important to you vs. which are supposed to 404, so we show you all the 404s we found on your site and let you decide which, if any, require your attention.
Markdown
To give a hover hint for a link, enter the hint after the link like [link] (example.com/bar.html "Bar is not about Foo").
Writing Speed
I took 4.5 hours to write the ~1k words in Pop Quiz: Falsifiable Hypotheses in 2015. But being a quiz made this easier to write than my original essays.
I took ~5 hours to write ~1.5k words in Rationalist’s Taboo in 2015.
Hypothesis: My writing speed is 300 words of final draft per hour.
Writing Stats
I seem to have written 348k words in 2017. Who would have thought?!
Essay = Causal models?
Test: Rationalist’s Taboo
Deep down inside of me
This section gives two examples of unfalsifiable statements that seem to give us some vague predictions.
Mum’s the Word
This one shows how to play Rationalist’s Taboo on The Dark Knight and Scarlett Johannsson so that you get clear predictions.
What do you say about this?!
Definition of Taboo. Aim of the game. Why it works - words vs predictions.
Taboo 101
How to play Taboo - Don’t use jargon. Block out adjectives and verbs. Unpack your categories.
No Place to Hide
Corollary: Everything is a question of fact now. Examples of places where you’re no longer entitled to your opinions.
My summary:
Hypothesis: Aim of the above essay: To show which kind of statement leads to no predictions and which one to clear predictions.
Taboo turns vague statements into clear ones.
How, though? A clear statement needs to predict some outcome for a variable. The idea is that simple words guarantee us clear predictions for the variables they talk about. So, using simple words is a surefire way to be precise, just like a working program tells you what will happen in a way that a hand-wavy high-level algorithm can’t.
Adjectives and verbs are the prime mischief-makers. For example, “best movie” doesn’t really give you clear predictions. So, replace them with simple words.
Unpack high-level labels so that you can see the actual variables involved. You may blank on “respiratory viral disease” but may recognize cold or influenza or measles by themselves.
Hypothesis: Simple words probably describe variables that are easy to observe or manipulate, like “car” or “tree” or “brown”, whereas “complex” words describe variables that are hard to observe or manipulate objectively, like “intelligent” or “best” or “hottest”. (Isn’t “best” a simple word?)
Corollary: You can judge the clarity of a statement by checking how many vague adjectives and verbs it has.
Corollary: You can make any statement falsifiable (by the definition of falsifiability viz it should have testable predictions). So, all the previously unfalsifiable statements people made are now falsifiable.
Overall summary: Aim - How to get testable predictions? An objectively testable prediction requires variables you can observe in an objective manner. “Objective” here means that humans can independently observe and agree on the outcome. Simple words are the ones that describe variables in a manner that we can all observe and agree on (and if we can’t, then we can drill down further; if I say a car is brown and you say it’s magenta, we can bring out the colour wheel because we agree with the names on that wheel).
Adjectives and verbs are what we commonly disagree on or cannot even observe. So, replace them with words whose values we can agree upon. High-level nouns too - replace them.
Corollary: We can never agree to disagree. We must agree to agree.
Hypothesis: An essay consists of causes and effects of a key variable, along with its definition in simple words.
comments powered by Disqus