11 April, 2016

A Meditation on Writing

[Epistemic Status: Uncertain]


Suppose you're consuming some fiction. It's prose is readable, it's ideas coherent. The plot isn't mind-blowing, but it's compelling. Say it's some maybe-supernatural mystery/thriller or something. In it, the story is set off by some local detectives receiving a tip from a Joe Everyman about some weird stuff happening in town. And lo behold, upon investigation, a frighteningly competent local neighborhood conspiracy is brought to light.

That's all well and good, nothing to write home about, maybe, but all good and well. But you're grabbing a drink from the fridge, as it goes, and you have a thought.

If this conspiracy is so competent, how did Joe Everyman even notice enough to give a tip?

It gets worse.

Suppose this Joe actually helps out the investigation. Maybe he contributes valuable input, some insight, a deduction or two, to this team of otherwise competent investigators. He even pulls a few tricks from his sleeve, gets them out of some sticky situations. Seems a bit Sue-ish, perhaps?

It gets worse.

Suppose this fiction wasn't just any fiction, but fanfiction.

You'd drop this poorly-thought-out, self-inserted crap right?

I probably would.

But. What if it turns out Joe isn't just an Everyman, but (say) a agent of a much larger, greater conspiracy, that maybe the local villains were rogue elements and he was just here to take them out, perhaps, and recruited the investigators to avoid getting his hands dirty.

Eliezer has aid some good stuff about noticing confusion. Those details weren't supposed to add up. It was a clue.

But no one kept reading that far.


There's something of an incompleteness theorem of writing, and it feels related to the Halting Problem and maybe Chaitin's Incompleteness Theorem.

(I mentioned something like in my last post, with those metamathematical curiosities.)

Given a story, it's impossible to tell if the incongruent details are there because the author is dumber than you or because he's smarter than you. Put less polemically, it's impossible to tell if the incongruencies are ever resolved until after they are actually resolved, analogous to being unable to tell if a program actually halts until to observe it halting.

Can we take this analogy farther?

There's an idea which has been baking in my head for a while, it's the concept of a 'reasonableness proof' and that an author must provide a reasonableness proof for everything that takes longer than a few steps of reasoning to justify. In the example above, the readers were expecting a reasonableness proof for the Joe Everyman character to have plot-relevant knowledge and abilities, and there wasn't one.

I've heard, quite a long while ago (so long that I don't even know where I heard it, but I suspect Yvain/Scott Alexander said it sometime) of a distinction between a 'series of suppostions' and a narrative, and the principle difference between the two is analogous to the difference between Peano Arithmetic, and taking all true statements about arithmetic as axioms, In one you can do mathematics, in the other is just a solid, uncompressed, uncomputable mess.

A better way of illuminating this is to say it looks like I just pushed the mysterious Joe Everyman into the premise, where everything required for the story to just function goes. There are many objections to doing this in more abstract domains. And I like applying that to stories.

It's one thing to say "What if there was mysterious and competent local conspiracy doing spooky things." and another to say "What if there was a mysterious and competent local conspiracy doing spooky things and this guys whose a total baddass and super-smart and awesome and figures them out and helps those other guys take them down.". One is parsimonious, simple, elegant even. The other is . . .  not.

Turns out I was really saying "What if there was a massive spooky conspiracy large enough that parts of it have fractured off into subconspiracies with opposed goals" which is more complex than the original, but simpler than the second while keeping a big part of it's content and adding more plot hooks.

So in a sense, these dramatically twists are strategically delaying reasonableness proofs until the narratively appropriate, I like doing this, even in my non-fiction writing. The biggest trouble with this is you layer too many mysteries on, without giving enough reasonableness proofs and your readers just think you're writing randomness or incoherence.

And like most things in life, the opposite sucks too, It'd be just a long series of expositions and characters doing things you have no interest in.


I like the metaphor between writing and math-ing. Premises as axioms, narratives as conjectures. The problem is language is inconsistent. Or it feels that way. Maybe it doesn't even have the epistemic machinery to be (in)consistent.

But let's restrict ourselves to the semantics of narratives, of stories. It seems we can render language constative by this, we could say something is true in the fictional world if, well, it's true in the story's world. And inversely for falsity. We could call something justifiable, iff there's a reasonableness proof of it.

So, in true logical fashion, we ask:

If a thing in a fictional world is justifiable, is a true (in the fictional world)?

It seems reasonable, but in most stories it probably fails to hold. In fact, personally, I'd go farther and argue that in every fictional world ever conceived, there is something justifiable that contradicts what happens in the story.

But that's part and parcel of being fictional.


In case it isn't clear what the model I'm sketching here is, I'm saying:

  • fictional worlds have premises, which is everything which is necessary (and maybe sufficient) for the story to hold
  • authors writing stories must provide reasonableness proofs, i.e. something in the story which justifies everything that isn't implied by the premises by more than a logical leap or two
  • stories are like conjectures, and you can say a story is inconsistent if we can find a reasonableness proof of something that contradicts 
  • you can call a world inconsistent if there's reasonableness proof that something both is and isn't true.

No comments:

Post a Comment