From 9b9e4a496560c77923cca43a3c1648f8bbad510f Mon Sep 17 00:00:00 2001 From: "M. Taylor Saotome-Westlake" Date: Sun, 28 Jan 2018 22:30:14 -0800 Subject: [PATCH] drafting "The Categories" and "Blegg Mode" --- content/drafts/blegg-mode.md | 12 ++++++------ ...were-made-for-man-in-order-to-make-predictions.md | 11 ++++++----- 2 files changed, 12 insertions(+), 11 deletions(-) diff --git a/content/drafts/blegg-mode.md b/content/drafts/blegg-mode.md index fbcd7bf..b05a2e2 100644 --- a/content/drafts/blegg-mode.md +++ b/content/drafts/blegg-mode.md @@ -6,20 +6,20 @@ Status: draft As part of a series—ah, Sequence—of [posts explaining the hidden Bayesian structure of language](https://www.lesserwrong.com/sequences/SGB7Y5WERh4skwtnb), Eliezer Yudkowsky [discusses](http://lesswrong.com/lw/nm/disguised_queries/) [a parable](http://lesswrong.com/lw/nn/neural_categories/) [about](http://lesswrong.com/lw/no/how_an_algorithm_feels_from_inside/) factory workers faced with the task of sorting objects which very strongly tend to _either_ be blue, egg-shaped, furry, flexible, opaque, luminescent, and vanadium-cored (categorized by the workers as "bleggs"), _or_ red, cube-shaped, smooth, hard, translucent, non-luminescent, and palladium-cored (categorized by the workers as "rubes"). -I want you to imagine that you're a worker in this factory, and occasionally, an object comes down the conveyor belt that's blue, _roughly_ egg-shaped, and furry, but also hard (unlike the typical blegg, which is slightly flexible to the touch). If such objects are extremely rare, you might not notice them at all—you'd quickly categorize each one as a _blegg_ and toss it in the blegg bin without a second thought. But as these unusual hard bleggs start to become more common, you get curious and take the time to examine one. +I want you to imagine that you're a worker in this factory, and occasionally, an object comes down the conveyor belt that's blue, _roughly_ egg-shaped, and furry, but also hard (unlike the typical blegg, which is slightly flexible to the touch). If such objects are extremely rare, you might not notice them at all—you'd quickly categorize each one as a _blegg_ and toss it in the blegg bin without a second thought. But as these unusual hard bleggs start to become more common, you notice them, get curious, and take the time to examine one. You make a startling discovery—the object was originally a smooth, hard red cube, of which someone had sanded down the corners to approximate an egg shape, and ironed on a layer of blue _faux_ fur. You show your work to Susan the Senior Sorter. "Wow," she says, "someone sure has gone to a lot of trouble to make these rubes look like bleggs!" -"Hold on," you say, "I'm not sure we should be disrespecting that effort by calling them _rubes_. [The categories were made for man, not man for the categories.](http://slatestarcodex.com/2014/11/21/the-categories-were-made-for-man-not-man-for-the-categories/) There's no rule of sorting saying that we should call them rubes, and there are plenty of rules of human decency saying that we should call them bleggs. And at a glance, they _look_ like bleggs—I mean, like the more-typical bleggs." +"Hold on," you say, "I'm not sure we should be disrespecting that effort by calling them _rubes_. [The categories were made for man, not man for the categories](http://slatestarcodex.com/2014/11/21/the-categories-were-made-for-man-not-man-for-the-categories/): there's no rule of sorting saying that we should call them rubes, and there are plenty of rules of human decency saying that we should call them bleggs. And at a glance, they _look_ like bleggs—I mean, like the more-typical bleggs." Susan rolls her eyes at you, but apparently doesn't care enough to argue about it, so the two of you agree to call the modified hard objects _adapted bleggs_ and get back to work. -Further investigation reveals that 85% of adapted bleggs don't glow in the dark and do contain palladium ore—both rube-typical chracteristics. Furthermore, the minority of adapted bleggs that glow and contain vanadium also seem more similar to typical bleggs in other ways: for example, by being somewhat more successfully egg-shaped and having more realistic _faux_ fur than the non-glowing, palladium-containing adapted bleggs. +Further investigation reveals that 90% of the adapted bleggs—like 98% of rubes, and like only 2% of non-adapted bleggs—contain fragments of palladium. -As the days go on, you find yourself taking notice of adapted bleggs—now that you're aware of their existence, they're not too hard to spot (although you have no way of knowing how many sucessfully "passing" adapted bleggs you've missed), and you need to take them to the sorting scanner so that you can put the palladium-containing ones in the palladium bin (formerly known as the _rube bin_). +As the days go on, you find yourself taking notice of adapted bleggs—now that you're aware of their existence, they're not too hard to spot (although you have no way of knowing how many sucessfully "passing" adapted bleggs you've missed), and you need to take them to the sorting scanner so that you can put the majority of palladium-containing ones in the palladium bin (formerly known as the _rube bin_). You notice that—despite having insisted on the neutral adjective _adapted_ rather than something perjorative like _counterfeit_ to descibe the modified objects—you don't really put them in the same mental category as bleggs: they seem to occupy a liminal third category in your internal ontology of sortable objects. -You ponder what this matter has taught you about the nature of categorization. (This job is so boring that you need to think about philosophy to keep your mind occupied while you sort.) You can think of two main lessons— +You ponder what this matter has taught you about the nature of categorization: what kind of structure does a population of objects need to exhibit in order for an efficient cognitive architecture to find it profitable to reify them as a distinct _category_ of object? (This job is so boring that you need to do philosophy of cognitive science to keep your mind occupied while you sort.) -First, \ No newline at end of file +After some thought, you conjecture that it must have something to do with having cheap-to-detect features that correlate with more-expensive-to-detect features that are decision-relevant with respect to the agent's goals. diff --git a/content/drafts/the-categories-were-made-for-man-in-order-to-make-predictions.md b/content/drafts/the-categories-were-made-for-man-in-order-to-make-predictions.md index 2574a6f..723dca3 100644 --- a/content/drafts/the-categories-were-made-for-man-in-order-to-make-predictions.md +++ b/content/drafts/the-categories-were-made-for-man-in-order-to-make-predictions.md @@ -10,7 +10,7 @@ Status: draft > > —_Distress_ by Greg Egan -In ["The Categories Were Made for Man, Not Man for the Categories"](http://slatestarcodex.com/2014/11/21/the-categories-were-made-for-man-not-man-for-the-categories/), the immortal Scott Alexander argues that proposed definitions of concepts aren't true or false in themselves, but rather can only be evaluated by their usefulness. Our finite minds being unable to cope with the unimaginable complexity of the raw physical universe, we group sufficiently similar things into the same category so that we can make similar predictions about them—but this requires not only a metric of "similarity", but also a notion of which predictions one cares about enough to notice, both of which are relative to some agent's perspective, rather than being inherent in the world itself. +In ["The Categories Were Made for Man, Not Man for the Categories"](http://slatestarcodex.com/2014/11/21/the-categories-were-made-for-man-not-man-for-the-categories/), the immortal Scott Alexander argues that proposed definitions of concepts aren't true or false in themselves, but rather can only be evaluated by their usefulness. Our finite minds being unable to cope with the unimaginable complexity of the raw physical universe, we group sufficiently similar things into the same category so that we can make similar [predictions](http://lesswrong.com/lw/i3/making_beliefs_pay_rent_in_anticipated_experiences) about them—but this requires not only a metric of "similarity", but also a notion of which predictions one cares about enough to notice, both of which are relative to some agent's perspective, rather than being inherent in the world itself. And so, Alexander explains, the ancient Hebrews weren't _wrong_ to classify whales as a type of _dag_ (typically translated as _fish_), even though modern biologists classify whales as mammals and not fish, because the ancient Hebrews were more interested in distinguishing which animals live in the water rather than which animals are phylogenetically related. Similarly, borders between countries are agreed upon for a variety of pragmatic reasons, and can be quite convoluted—while there may often be some "obvious" geographic or cultural Schelling points anchoring these decisions, there's not going to be any intrinsic, eternal fact of the matter as to where one country starts and another begins. @@ -48,7 +48,7 @@ This is true in a tautological sense: if you deliberately gerrymander your categ But it's not very interesting to people like rationalists—although apparently not all people who _self-identify_ as rationalists—who want to use concepts to _describe reality_. -Alexander gives an account of a woman whose ability to function at work was being disrupted by obsessive-compulsive fears of leaving her hair dryer on at home, whose problems were solved by the simple expediency of taking the hair dryer with her when leaving the house. Given that it _worked_ to resolve her distress, we shouldn't care that this isn't how problems that are categorized as _obsessive-compulsive disorder_ are "supposed" to be treated, and Alexander argues that the same should go for accepting transgender identity claims: if it _works_ for resolving people's gender dysphoria, why not? +Alexander gives an account of a woman whose ability to function at her job was being disrupted by obsessive-compulsive fears of leaving her hair dryer on at home, whose problems were solved by the simple expediency of taking the hair dryer with her when leaving the house. Given that it _worked_ to resolve her distress, we shouldn't care that this isn't how problems that are categorized as _obsessive-compulsive disorder_ are "supposed" to be treated, and Alexander argues that the same should go for accepting transgender identity claims: if it _works_ for resolving people's gender dysphoria, why not? The problem is that there are _significant disanalogies_ between leaving a hair dryer in the front seat of one's car, and collectively agreeing that gender should be defined on the basis of self-identity. Most significantly: the former has no appreciable effects on anyone but the person themselves; the latter affects _everyone who wants to use language to categorize humans by sex_. The words _man_ and _woman_ [are top-20 nouns](https://en.wikipedia.org/wiki/Most_common_words_in_English#Nouns)! People need those nouns to describe their experiences! @@ -66,9 +66,10 @@ Well, no. But the point is that this is an _empirical_ argument for why successf Another factor affecting the degree to which trans people form a more natural category with their identified gender than their natal physiological sex is the nature of transgenderedness itself. If gender dysphoria is caused by a brain-restricted intersex condition, such that trans people's psychology is much more typical of the other physiological sex—if the "woman trapped in a man's body" trope is basically accurate—that would tend to weigh in favor of accepting transgender identity claims: trans women would be "coming from the same place" as cis women in a real psychological sense, despite their natal physiology. -On the other hand, if gender dysphoria is caused by something else, that would tend to weigh against accepting transgender identity claims: however strongly felt trans people's _subjective_ sense of gender identity might be, if the mechanism underlying that feeling actually has nothing in particular in common with people of the identified-with sex, it becomes more tempting to classify the subjective sense of gender identity as an illusion, rather than the joint in reality around which everyone needs to carve their gender categories. - -Of course, the phrasing _If gender dysphoria is caused by ..._ implies that we're considering _gender dysphoria_ as one category to reason about homogenously. But different people might want to transition for very different underlying psychological reasons. (Compare how many different pathogens can all cause the same symptoms of coughing and sneezing.) +On the other hand, if gender dysphoria is caused by something else, that would tend to weigh against accepting transgender identity claims: however strongly felt trans people's _subjective_ sense of gender identity might be, if the mechanism underlying that feeling actually has nothing in particular in common with people of the identified-with sex, it becomes relatively more tempting to classify the subjective sense of gender identity as an illusion, rather than the joint in reality around which everyone needs to carve their gender categories. +Of course, the phrasing _If gender dysphoria is caused by ..._ implies that we're considering _gender dysphoria_ as one category to reason about homogenously. But different people might want to transition for very different underlying psychological reasons. What categories we use may not be a question of simple fact that we can get wrong, but if, empirically, there happens to be a sufficiently robust statistical structure to the simple facts of the cases—if some people want to transition for reason _A_ and tend to have traits _W_ and _X_, but others want to transition for reason _B_ and have traits _Y_ and _Z_—then aspiring epistemic rationalists may find it useful to distinguish multiple, distinct psychological conditions that both happen to cause gender dysphoria as a symptom. +Analogously, in medicine, many different pathogens can cause the same symptoms (_e.g._, coughing and sneezing), but doctors care about distinguishing different illnesses by etiology, not just symptoms, because distinct physical mechanisms give rise to distinct treatment decisions, either immediately (_e.g._, a bacterial illness will respond to antibiotics, but a viral one won't) or in principle (_e.g._, today's treatments might be equally effective against two different species of bacteria, but future drugs might work better on one or the other). +_As it happens_, (I claim) the evidence that \ No newline at end of file -- 2.17.1