Less Wrong is a community blog devoted to refining the art of human rationality. Please visit our About page for more information.

steven0461 comments on Rationalist Fiction - Less Wrong

27 Post author: Eliezer_Yudkowsky 19 March 2009 08:22AM

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Comments (189)

You are viewing a single comment's thread. Show more comments above.

Comment author: Sideways 19 March 2009 08:53:22PM 9 points [-]

In defense of Sherlock Holmes:

The typical Sherlock Holmes story has Holmes perform twice. First he impresses his client with a seemingly impossible deduction; then he uses another deduction to solve the mystery. Watson or the client convince Holmes to explain the first deduction, which gives the reader the template Holmes will use for the second (likely inferences from small details). The data that Holmes uses to make the second deduction are in the text and available to the reader--the reader's challenge is to make Holmes's inference in advance.

Holmes himself attributes his success to observation, not rationality. (There's a startling passage in A Study In Scarlet where Holmes tells Watson that he can't be bothered to remember that the sun orbits the earth! Visit the link and search for 'Copernican Theory' in the full text for the passage.) The Sherlock Holmes stories are intended to be exercises in attention to detail, which is surely a useful skill for a rationalist.

Comment author: steven0461 19 March 2009 10:18:04PM 75 points [-]

When you have eliminated the impossible, whatever remains is often more improbable than your having made a mistake in one of your impossibility proofs.

Comment author: CarlShulman 22 March 2009 04:48:18AM 24 points [-]

Beautiful comment, but I'd add that whatever remains of the hypotheses you considered is often more improbable than your having missed an unconsidered alternative.

Comment author: Jack 21 December 2009 12:50:13AM 10 points [-]

I just stumbled across this and felt this comment and the one above it were worth reminding everyone of in light the Knox case discussion. Way too many of our discussions have involved trying to come up with accounts of the crime that make sense of all the evidence. In retrospect I would labels such discussions as fun, but unhelpful.

Comment author: NancyLebovitz 19 February 2011 09:45:07PM 12 points [-]

This reminds me of a bit in the Illuminatus! trilogy-- there was a man who had filing cabinets full of information about the Kennedy assassination. [1]

He kept hoping that he'd find one more piece of information which would make sense of everything he'd accumulated, little realizing that most of what he had was people getting things wrong and covering their asses.

[1] Once upon a time, it was normal to store information in filing cabinets, and there was only one Kennedy assassination.

Comment author: Apprentice 08 June 2011 09:46:40PM 5 points [-]

Sounds like me and my PhD project.

Comment author: NancyLebovitz 08 June 2011 09:58:23PM 3 points [-]

If you don't mind, what was the subject?

Comment author: MBlume 21 March 2012 09:48:41PM 10 points [-]

I reject that entirely. The impossible often has a kind of integrity to it which the merely improbable lacks. How often have you been presented with an apparently rational explanation of something which works in all respects other than one, which is just that it is hopelessly improbable? Your instinct is to say, ‘Yes, but he or she simply wouldn’t do that.’

-- Dirk Gently

Comment author: lmm 07 January 2014 11:28:17PM 2 points [-]

I view Dirk Gently as a kind of wonderfully effective strawman, and his stories were a great aid to realizing I was an atheist, because at first he seems correct: surely, rather than a "localized meteorological phenomenon", it makes more sense that the guy who's been rained on for 14 straight years is some kind of rain god.

And then you think about what would happen in the real world, and realize that no, even if someone had been rained on for 14 years straight, I would not believe that they were a rain god. Because rain gods are actually impossible.

That part hit me like a punch in the gut.

Comment author: TheOtherDave 08 January 2014 04:01:09AM 5 points [-]

In the real world, these are mostly just games we play with words.

Someone who has been rained on for 14 years straight has an extremely surprising property.

The label we assign that property matters a little, since it affects our subsequent behavior with respect to it. If I call it "rain god" I may be more inclined to worship it; if I label it a "localized meteorological phenomenon" I might be more inclined to study it using the techniques of meteorology; if I label it an extremely unlikely coincidence I might be more inclined not to study it at all; if I label it the work of pranksters with advanced technology I might be more inclined to look for pranksters, etc.

Etc.

But other things matter far more.

Do they have any other equally unlikely observable attributes, for example?
Did anything equally unlikely occur 14 years ago?

Etc.

Worrying overmuch about labels can distract us from actually observing what's in front of us.

Comment author: VAuroch 08 January 2014 04:47:42AM 0 points [-]

So it wouldn't be possible to convince you that 2+2=3? No matter the evidence?

If someone claimed to be a rain god, or was credibly claimed to be a rain god based on previous evidence, and tested this by going through an EMP, stripping, generally removing any plausible way technological means could be associated with them, then being transported while in a medically-induced coma to a series of destinations not disclosed to them in advance in large deserts, and at all times was directly under, in, or above rainclouds, defying all meteorological patterns predicted by the best models just in advance of the trip, I find it hard to see how you could reasonably fail to assign significant probability to a model which made the same predictions as "this person is a rain god".

Comment author: lmm 10 January 2014 12:38:07PM 0 points [-]

It'd be possible, but it would take more evidence than someone having been rained on for 14 years.

If you're talking about models and predictions you've already made the relevant leap, IMO. Even if you're calling the person a "god", you're still taking a fundamentally naturalistic approach; you're not assuming basic mental entities, you're not worshiping.

Comment author: VAuroch 11 January 2014 01:26:21AM -1 points [-]

Calling someone a rain god is making the prediction "If I worship this person, rain will occur at the times I need it more often than it would if I did not worship this person." Worship doesn't stop being worship just because it works.

Comment author: hyporational 08 January 2014 06:39:52AM 0 points [-]

Where does personal insanity become a factor in your probability estimates?

Comment author: VAuroch 09 January 2014 04:37:11AM 1 point [-]

In some sense, basically everywhere there is a very-low or very-high probability belief, since obviously I can't be more confident in any belief than I can be in the reliableness of my system of reasoning. I definitely consider this when I'm evaluating the proper strength of nearly-certain beliefs. In another sense, almost nowhere.

I don't know exactly how confident I should be in my sanity, except that the probability of insanity is small. Also, I'm not confident there would be any evidence distinguishing 'sane and rational' from 'insane but apparently rational'. I model a logical-insane VAuroch as being like the anti-inductors; following different rules which, according to their own standards, are self-consistent.

Since I can't determine how to quantify it, my response has been to treat all other beliefs as conditioned on "my reasoning process is basically sound", which makes a fair number of my beliefs having tacit probability 1; if I find reason to question any of these beliefs, I will have to rederive every belief from the original evidence as much as possible, because it's exposed a significant flaw in the means by which I determine what beliefs to hold. Largely this consists of mathematical proofs, but also things like "there is not currently a flying green elephant in this room" and "an extant rain god is mutually incompatible with reductionism".

Comment author: ialdabaoth 09 January 2014 04:52:24AM 0 points [-]

Since I can't determine how to quantify it, my response has been to treat all other beliefs as conditioned on "my reasoning process is basically sound", which makes a fair number of my beliefs having tacit probability 1; if I find reason to question any of these beliefs, I will have to rederive every belief from the original evidence as much as possible, because it's exposed a significant flaw in the means by which I determine what beliefs to hold. Largely this consists of mathematical proofs, but also things like "there is not currently a flying green elephant in this room" and "an extant rain god is mutually incompatible with reductionism".

This is an amazingly apt description of the mind-state that Robert Anton Wilson called "Chapel Perilous".

Comment author: VAuroch 09 January 2014 05:19:52AM *  -1 points [-]

It is interesting that you think so, but I can't make head or tail of his description of the state, and other descriptions don't bear any particular resemblance to the state of mind I describe.

My position on the matter boils down to "All my beliefs may be unjustified, but until I have evidence suggesting they are, I should provisionally assume the opposite, because worrying about it is counterproductive."