Let There Be Light

Author: Alicorn. Link to original: http://lesswrong.com/lw/1xq/let_there_be_light/ (English).
Tags: lesswrong, Rationality Submitted by Remlin 10.01.2013. Public material.

Translations of this material:

into Tatar: Перевод "Let There Be Light ". Translation is not started yet.
Submitted for translation by hamidullakhodja 28.11.2017
into Russian: Да будет свет. Translation complete.
Submitted for translation by Remlin 10.01.2013 Published 4 years, 12 months ago.

Text

You can start from psych studies, personality tests, and feedback from people you know when you're learning about yourself. Then you can throw out the stuff that sounds off, keep what sounds good, and move on.

You may find your understanding of this post significantly improved if you read the first story from Seven Shiny Stories.

Where do you get your priors, when you start modeling yourself seriously instead of doing it by halfhearted intuition?

Well, one thing's for sure: not with the caliber of introspection you're most likely starting with. If you've spent any time on this site at all, you know people are riddled with biases and mechanisms for self-deception that systematically confound us about who we are. ("I'm splendid and brilliant! The last five hundred times I did non-splendid non-brilliant things were outrageous flukes!") Humans suck at most things, and obeying the edict "Know thyself!" is not a special case.

The outside view has gotten a bit of a bad rap, but I'm going to defend it - as a jumping-off point, anyway - when I fill our luminosity toolbox. There's a major body of literature designed to figure out just what the hell happens inside our skulls: it's called psychology, and they have a rather impressive track record. For instance, learning about heuristics and biases may let you detect them in action in yourself. I can often tell when I'm about to be subject to the bystander effect ("There is someone sitting in the middle of the road. Should I call 911? I mean, she's sitting up and everything and there are non-alarmed people looking at her - but gosh, I probably don't look alarmed either..."), have made some progress in reducing the extent to which I generalize from one example ("How are you not all driven insane by the spatters of oil all over the stove?!"), and am suspicious when I think I might be above average in some way and have no hard data to back it up ("Now I can be confident that I am in fact good at this sort of problem: I answered all of these questions and most people can't, according to someone who has no motivation to lie!"). Now, even if you are a standard psych study subject, of course you aren't going to align with every psychological finding ever. They don't even align perfectly with each other. But - controlling for some huge, obvious factors, like if you have a mental illness - it's a good place to start.

For narrowing things down beyond what's been turned up as typical human reactions to things, you can try personality tests like Myers-Briggs or Big Five. These are not fantastically reliable sources. However, some of them have some ability to track with some parts of reality. Accordingly, saturate with all the test data you can stand. Filter it for what sounds right ("gosh, I guess I do tend to be rather bothered by things out of place in my environment, compared to others") and dump the rest ("huh? I'm not open to experience at all! I won't even try escargot!") - these are rough, first-approximation priors, not posteriors you should actually act on, and you can afford a clumsy process this early in the game. While you're at it, give some thought to your intelligence types, categorize your love language1 - anything that carves up person-space and puts you in a bit of it.

Additionally, if you have honest friends or relatives, you can ask for their help. Note that even honest ones will probably have a rosy picture of you: they can stand to be around you, so they probably aren't paying excruciatingly close attention to your flaws, and may exaggerate the importance of your virtues relative to a neutral observer's hypothetical opinion. They also aren't around you all the time, which will constrict the circumstances in which their model is tested and skew it towards whatever influence their own presence has on you. Their outside perspective is, however, still valuable.

(Tips on getting friends/family to provide feedback: I find musing aloud about myself in an obviously tentative manner to be fairly useful at eliciting some domain-specific input. Some of my friends I can ask point-blank, although it helps to ask about specific situations ("Do you think I'm just tired?" "Was I over the line back there?") rather than general traits that feel more judgmental to discuss ("Am I a jerk?" "Do I use people?"). When you communicate in text and keep logs, you can send people pastes of entire conversations (when this is permissible to your original interlocutor) and ask what your consultant thinks of that. If you do not remember some event, or are willing to pretend not to remember the event, then you can get whoever was with you at the time to recount it from their perspective - this process will automatically paint what you did during the event in the light of outside scrutiny.)

If during your prior-hunting something turns up that seems wrong to you, whether it's a whole test result or some specific supposed feature of people in a group that seems otherwise generally fitting, that's great! Now you can rule something out. Think: what makes the model wrong? When have you done something that falsified it? ("That one time last week" is more promising than "back in eighty-nine I think it might have been January".) What are the smallest things you could change to make it sit right? ("Change the word "rapid" to "meticulous" and that's me to a tee!") If it helps, take in the information you gather in small chunks. That way you can inspect them one at a time, instead of only holistically accepting or rejecting what a given test tells you.

If something sounds right to you, that's also great! Ask: what predictions does this idea let you make about your cognition and behavior? ("Should you happen to meet a tall, dark stranger, you will make rapid assumptions about his character based on his body language.") How could you test them, and refine the model? (Where do the tall, dark strangers hang out?) If you've behaved in ways inconsistent with this model in the past, what exceptions to the rule does that imply and how can you most concisely, Occam-esque-ly summarize them? ("That one tall, dark stranger was wearing a very cool t-shirt which occluded posture data.")

Nota bene: you may be tempted to throw out things because they sound bad ("I can't be a narcissist! That wouldn't be in keeping with the story I tell about myself!"), rather than because they sound wrong, and to keep things because they sound good ("ooh! I'm funny and smart!"), rather than because they sound right. Recite the Litany of Tarski a few times, if that helps: if you have a trait, you desire to believe that you have the trait. If you do not have a trait, you desire to believe that you do not have the trait. May you not become attached to beliefs you may not want. If you have bad features, knowing about them won't make them worse - and might let you fix, work around, or mitigate them. If you lack good features, deluding yourself about them won't make them appear - and might cost you opportunities to develop them for real. If you can't answer the questions "when have you done something that falsified this model?" or "list some examples of times when you've behaved in accordance with this model" - second guess. Try again. Think harder. You are not guaranteed to be right, and being right should be the aim here.

1It looks cheesy, but I've found it remarkably useful as a first-pass approximation of how to deal with people when I've gotten them to answer the question.