Level 3 Intelligent Characters and the Methods of Rationality
One of the most heartwarming compliments I’ve received on HPMOR was an image on a confessions site, I don’t recall the exact location.
The caption of the image was: “My thesis advisor thinks I’m brilliant. I don’t dare tell him that I’m just doing what I think HE would do.”
The image was a boy with a scar on his forehead, snapping his fingers.
(To whoever put that up: if you can imagine HJPEV well enough to know what experiments he would do and thereby impress your thesis advisor, then Vinge’s Principle says you are at least as smart as the Harry Potter who lives in your head. It’s not like your model of him is running on someone else’s brain.)
A year before I started on HPMOR, I defined the term “rationalist fiction” to refer to a very few existing works like The World of Null-A by A. E. van Vogt. It had no connotation of fanfiction at the time (March 2009); I had only written original stories at that point in my writing career.
In 2009, I said as follows:
But when you look at what Sherlock Holmes does - you can’t go out and do it at home. Sherlock Holmes is not really operating by any sort of reproducible method. He is operating by magically finding the right clues and carrying out magically correct complicated chains of deduction. Maybe it’s just me, but it seems to me that reading Sherlock Holmes does not inspire you to go and do likewise. Holmes is a mutant superhero. And even if you did try to imitate him, it would never work in real life.
Contrast to A. E. van Vogt’s Null-A novels, starting with The World of Null-A. Now let it first be admitted that van Vogt had a number of flaws as an author. With that said, it is probably a historical fact about my causal origins, that the Null-A books had an impact on my mind that I didn’t even realize until years later. It’s not the sort of book that I read over and over again, I read it and then put it down, but this is where I was first exposed to such concepts as “The map is not the territory” and “rose1 is not rose2”.
Null-A stands for “Non-Aristotelian”, and the premise of the ficton is that studying Korzybski’s General Semantics makes you a superhero. Let’s not really go into that part. But in the Null-A ficton:
1) The protagonist, Gilbert Gosseyn, is not a mutant. He has studied rationality techniques that have been systematized and are used by other members of his society, not just him.
2) Van Vogt tells us what (some of) these principles are, rather than leaving them mysteriously blank—-we can’t be Gilbert Gosseyn, but we can at least use some of this stuff.
3) Van Vogt conveys the experience, shows Gosseyn in action using the principles, rather than leaving them to triumphant explanation afterward. We are put into Gosseyn’s shoes at the moment of his e.g. making a conscious distinction between two different things referred to by the same name.
What is this mysterious stuff called by the name rationality? Why, it is exactly those techniques of good thinking which are not incommunicable, not private, not forever mysterious. They are the ways of good thinking that operate by understandable rules, which therefore can be said out loud, depicted in detail, and explained to other people.
If you’re exposed to good thinking in any recognizable form, you ought to be learning it at least a little. If you can truly recognize good features of a thought process when you read about it, you should stand a better chance of reproducing them yourself later.
In the theory of Artificial Intelligence there is a duality between recognizing good solutions and inventing good solutions. In the limit of unbounded computing power, to invent a good solution, we just take a good-solution-recognizer and run it over all possible inputs. In real life we have bounded computing power, but that doesn’t change the structure of the problem: the ability to recognize a good thought is inherently linked to the ability to invent good thoughts.
If you can recognize a set of thoughts as intelligent, then in the limit of unbounded computing power you can be intelligent; just search all possible thoughts. And even in the real world of bounded computing power there’s still a connection, even if it’s not an identity. Seeing good examples should train your recognition, make you faster to recognize things like that.
So if you’re not becoming more intelligent from reading a character’s viewpoint, not even a little, then the author must not be showing real intelligence.
Sure, a real-world genius will have talents you can’t absorb just by watching them think. But neither are realistic genius-level thoughts made entirely out of opaque function calls. Reading Surely You’re Joking, Mr. Feynman didn’t turn all the readers into Nobel laureate physicists, but a whole lot of readers managed to absorb something from it, even if just a little.
So any supposedly smart character whose smartness readers can’t absorb, not even a little? That character’s thoughts, at least the ones you’re shown, must not actually be smart.
And if you can learn some skill of ‘intelligence’ by watching a literary character do it… why, you might as well call that one of the methods of rationality. It’s communicable; it’s not a mutant superpower.
Everything I said about explaining science in your fiction, applies also to explaining cognitive science, or any technique of rationality. It must be truly relevant (either the plot was built around it, or it fit naturally to the plot) so that the explanation is part of the plot tension rather than a diversion from it. You must know how to explain it in your own terms, paring away all jargon and standard concepts. In many cases (though not all) the standard names should disappear, leaving only the idea’s use.
The result, if this skill roll succeeds, is a Level 3 Intelligent Character whose underlying thought processes glow with intelligence even apart from their Level 2 strokes of genius, so that readers feel their souls sway to the rhythm of surprisingly good thinking, and find themselves able to imitate those characters at least a little. It’s all about reproducible patterns of good thought, which in turn challenges you, the author, to know explicitly which features mark the character’s good thinking as good (whether or not you say so in the writing).
And similarly, when your character makes a mistake (preferably by executing a well-intentioned attempt at thinking not quite well enough) it will help to know explicitly what fallacy or bias or antipattern has led to that mistake, so that afterward, when your character thinks about how they will never make that mistake again, the reader learns the same lesson explicitly.
The readers who intensely loved the intelligence of the characters in HPMOR weren’t being impressed by the rocket-broomstick or other feats of Level 2 Intelligence. They were impressed by the detail on how Harry thinks through to the solutions he uses, the heuristics that Harry used to come up with his answer, the alternatives he considered and discarded. They felt themselves learning how to think better by watching Harry think, and from watching Harry himself learn to think better. Scroll through the latest reviews of HPMOR, and you’ll find a review saying just that before you’ve looked very far.
This level of character intelligence cannot be faked at all, by any literary artifice.
You cannot do it by namedropping cog-sci terminology, trying to signal that you are part of the rationalist ingroup, because reading those phrases will not teach anyone how to think. Even if they look up the phrase ‘planning fallacy’ online, you haven’t taught them to think from within your own writing, you have not shown-not-told intelligence and thereby sneezed it onto them.
You cannot take the standardized Deep Wisdom of your surrounding culture and make your characters rehearse it. What leaps into your mind as a cached thought for Deep Wisdom, even if you think that Deep Wisdom unknown to the laity, is nonetheless the equivalent of writing vampires who hiss and drink blood instead of asking yourself what you would do as a vampire. This isn’t to say that you have to invent your own version of the planning fallacy, but when the character thinks about the planning fallacy it should not be exactly the same thought that Harry thinks in HPMOR. For you, now, that is the equivalent of writing about vampires that hiss and drink blood. To say what all your friends are thinking or what you remember someone else writing, even if you think the reader shouldn’t have heard it before, will have a sense of staleness about it. To recognize this staleness and think past it, to a less rehearsed form that isn’t stale, involves a function call to Originality.
You cannot create recognizable, learnable intelligence by declaring that your characters use a technique and therefore win. Level 3 intelligence exists on the level of paragraphs, not whole plots. It is in how your character arrives at the next thought shown one paragraph later. The ultimate story outcome doesn’t matter to how well that cognitive algorithm would work in real life or whether the reader learned anything by reading it. Deciding that a character will win using some technique is telling the reader about intelligence, not showing it to them; it is praising the skill, but not yet explaining it. This isn’t to say that your characters can never win when they use techniques. It’s just that the part where the plot says they win doesn’t constitute part of the difficult work of showing good thinking (though how a victory plays out can be important to the illustration).
You can learn to write Level 3 Intelligent characters by living your own life well, learning about how to think and trying to explain what you’ve learned to others, studying the relevant sciences, seeking out techniques others have systematized and applying them to your own life. Whatever communicable cognitive skills you learn to the level of applying them to your own life and experience, you can lend a character one of those skills to use in their experience, and have the character think in a way that readers will absorb.
That is the great secret of writing viscerally intelligent characters, and it is just what it should be; that is all.