= 27d567bc2d48d9f40e9ccc20ea370b15)
Erebus comments on Rationality Lessons Learned from Irrational Adventures in Romance - Less Wrong
You are viewing a comment permalink. View the original post to see all comments and the full post content.
= 783df68a0f980790206b9ea87794c5b6)
Loading…= b715d02e85f7b3b4ae65ca3d3e450868)
Subscribe to RSS Feed
Powered by Reddit
= f037147d6e6c911a85753b9abdedda8d)
You are viewing a comment permalink. View the original post to see all comments and the full post content.
Comments (609)
Understanding signaling in communication is almost as basic as understanding the difference between the map and the territory.
A choice of words always contains an element of signaling. Generalizing statements are not always made in order to describe the territory with a simpler map, they are also made in order to signal that the exceptions from the general case are not worth mentioning. This element of signaling is also present, even if the generalization is made out of a simple desire to not "waste space" - indeed the exceptional cases were not mentioned! Thus a sweeping generalization is evidence for the proposition that the speaker doesn't consider the exceptions to the stated general rule worth much (an upper bound is the trouble of mentioning them). And when dealing with matters of personal identity, not all explanations for the small worth of the set of exceptional people are as charitable as a supposedly small size of the set.
Certainly.
However, the simple truth is that communication becomes positively impossible if 'sweeping generalizations' at some level are not made. Is this a trade-off? Sure. But I for one do not find it exceedingly difficult to treat all broad-category generalizations as simulacra representing the whole body. Just like how there's probably not a single person in politics who agrees with the entirety of the DNC or the GOP's platforms, discussing those platforms is still relevant for a reason.
And political identity is arguably one of the most flame-susceptible category of that available for discourse nowadays. So that's saying something significant here.
True but misleading. One should seek to avoid eliminating relevant meaning in the process of making those generalizations.
If you say "Men are sexually attracted to women" and your intended meaning is "this is true enough often enough to serve as a reliable guide to male behavior", then when someone points out that homosexual men and asexual men exist, the fact that those groups are minorities doesn't change the fact that you were imprecise in misleading ways, even if you didn't explicitly say "always". In addition, the unspoken implications you take out of the the statement (which could be nearly anything depending on what you're talking about) may be apparent but not agreeable to the listener, which is quite relevant if you're depending upon those to support your argument downstream.
So yes, make generalizations, but make good, accurate generalizations with appropriate scope limitations. And try to make the implications you perceive explicit.
(Formatting tip: you need to add two spaces at the end of the previous line to get lesswrong's commenting markup language to "<br>"/"\n". Two newlines will "<p>".)
I follow the convention of thinking that provisos are somwhere betwee standard deviation or significant digits. When someone adds that proviso "asexual/homosexual" -- they are changing the relevant level of precision necessary to the conversation.
For example; if I say "Men and women get married because they love each other", then the fact that some men/women don't marry, or the fact that intersex people aren't necessarily men or women, or the fact that GLBT people who marry are also likely to do so because of love, or the fact that some marriages are loveless is only a distraction to the conversation at hand.
While this seems like a trivial item for a single statement, the thing about this is that such provisos propagate across all dependent statements, meaning that the informational value of all dependent statements is reduced by each such proviso made.
Consider the difference in meaning between "Men and women marry each other because they love each other" and "Men/women/intersex individuals and other men/women/intersex individuals may or may not marry one another in groups as small as two with no upper bound for reasons that can vary depending on the situation."
This is, granted, an extreme example (reductio absurdum) but I make it to demonstrate the value of keeping in mind your threshold of significance when making a statement. Sometimes, as counterintuitively as it may seem, less accurate statements are less misleading.
I'm reasonably confident that most intersex people are either men or women. You meant genderqueer.
It's tough to get exact numbers on the rate of intersex individuals per thousand, but I do know that the number of intersex individuals I've met and known for some time is far higher than that rate. No, I did not mean "genderqueer". This would be what you might call "too many digits beyond what's significant."
Or meant to distinguish males, females and intersex persons rather than men, women and intersex persons.
Oooh, perfect example! Because this is probably still not true for a plurality, if not majority of humanity, and it used to be little more than a perk if it occurred in a marriage. For most of human history and for much of humanity today, marriage is more like a business relationship, corporate merger, pragmatic economic decision...
If you confine your statement to Westerners, and especially middle-to-upper class ones, and those who live in societies strongly modelled on the same pattern (urban Chinese often yes; rural Chinese often no) then you are dealing with an acceptable level of accurate to be relatively unobjectionable.
Do you want to try again?
[...]
My statement wasn't ever meant to be representative of the whole. That should have been obvious. If I'd said "only for love" then that'd be a valid objection. As it stands, I have no such problem. Generalizations that are useful for a context need not be without exception or even universally comprehensive.
People in the past or in other cultures are irrelevant to me when discussing social habits I am familiar with.
So, no. My statement is fine as is. Did I leave out a great heaping swath of precisions, provisos, and details? Absolutely!! -- but that was the point from the outset.
And you wouldn't hear a peep out of me if it wasn't depressingly common to see people couch advice, theories and other mental-model-of-the-world stuff in such terms, giving no obvious sign that they've thought about the distinction between "speaking to a specific audience" and just speaking with the assumption that the listeners fit their relatively vague preconception of who they talk to, rather than about.
It's far from clear when an Anglophonic Western man says "Men and women marry each other for romantic love" that he is cognizant of the distinction. After all, that's his default context, other possibilities are barely even mentioned in his expected cultural background (let alone presented as normal), and unless he has much overt contact with people for whom that's not the case, the odds are pretty good it's a thing-over-there, done by some outgroup about whom he knows rather little.
It may not be terribly important if he's just talking among a peer group of like folks, but when he's got access to a wide and relatively unknown audience (it could be anyone reading), and he's trying to frame it in terms of general information about "how people work", it's usually a safe bet he just didn't think about how his own norms influence his advice, and hence how applicable it might be to even, say, an English-speaking, technically-trained man in India (where arranged marriages for purposes other than romantic love are still pretty standard).
Sometimes people on this site even take norms like that and try to infer over all of human evolution. So yeah -- this is not an unreasonable thing to question.
Can you rephrase this for me? It's not parsing my language-interpreter.
Certainly. Arguably, for the majority of cases it's not even relevant whether he is or isn't. In all likelihood whoever he is talking to also shares that set -- as you said, it's his "default context". Now, yes, absolutely failing to recognize that one's default context is not the sole available context can be a significant problem. But that really isn't relevant to the topic of my assertions about cognitive burden per statement of equivalent informational value and the relevance of said burden to knowing when generalizing trivial elements of a statement is a net gain rather than net loss.
You know, after years of making daily calls to workers in India (I do corporate sysadmin work, for a number of various corporations) -- I still have absolutely no clue beyond the vaguest notions gleaned from the "idiot box" (TV, but at least I mean PBS-ish) about the cultural contexts of a modern urban Indian person. I really do feel like I understand more about the unspoken assumptions of Amazonian tribesmen than I do about Indian people.
I do, however, find it both insulting when my offshores co-workers think they can slip insults by me through such expedients as telling me to "do the needful" in a particular tone, but I digress.
Absolutely not an unreasonable thing to question, since any norm not empirically validated to exist in other monkeys (I am of the belief that all modern primates qualify monocladistically as monkeys) is simply not viable material for Evo-Psych theories without significant and rigorous documentation.
By the way, I just made an inaccurate statement for the purposes of making the statement less misleading, as I previously asserted. It has to do with my use of the term "empirically" -- I follow the thinking of Poplerian falsificationism which, while similar to empiricism, does not suffer from the problem of induction. While this one instance is trivial -- keeping up that level of technicality quickly turns casual conversation into cited, researched, thesis papers. And it's just plain impossible to always communicate at that level; ergo, devoting actual thought and consideration to building a rational heuristic for when generalization / inaccuracy is acceptable is a necessary part of the toolkit. Which is what I was saying from the outset.
No, they are pointing out that in order to apply to a case they are interested in, the conversation must be made more precise.
The last one isn't a distraction, it's a counterexample. If you want to meaningfully say that men and women marry out of love, you must implicitly claim that loveless marriages are a small minority. If someone says, "A significant number of of marriages are loveless," they aren't trying to get you to add a trivializing proviso. They're saying that your generalization is false.
This isn't a reductio, it's a strawman. When you add provisos to a statement that is really nontrivial, you do not turn "generally" into "may or may not." You turn "always" into "generally", or "generally" into "in the majority of cases".
In any case, what about "People who marry generally do so out of love?" This retains the substance of the original statement while incorporating the provisos. All that is gained is real clarity. All that is lost is fake clarity. (And if enough people are found who marry for other reasons, it is false.)
I want you to understand that you just agreed with me while appending the word "No" to the beginning of your sentence. This is... a less than positive indicator as to whether I am being understood.
The statement doesn't allow for counterexamples because it's a statement of fact, at bare minimum: the fact is that men and women do marry because they love each other. Other shit happens too, but that itself is a factual statement. Its informational value as a statement can only be derived from within the text of a given conversation.
That doesn't follow. Where do you get this necessity of implication from? Certainly not from the principle I'm espousing here. (Note: "A small minority" is a different statement from "a minority". In several cities in the US, whites are a minority. And yet the second-order simulacrum of those populations would still be a white person -- because whites, while a minority, are the plurality [largest minority].)
If and only if you meant "always" in the first place and want to be less than perfectly accurate. "In the majority of cases" is an inaccurate method of expressing how S-O S's work -- as I mentioned above, with "the largest minority" being the representative entity of the body. So you'd be better able to most accurately express the situation by stating that X happens Y percent of the time, but that simply isn't language used in ordinary discourse.
That the statement can be revised in this manner does not obviate the example I was pointing to with the previous example. I used an explicit reductio ad absurdum to make the mechanism explicit. From zero to one hundred, as it were.
In a more 'realistic' example for your revision: what is meant by "generally"? What is meant by "love"? What is meant by "people who marry"? These are all imprecise statements. Is "generally" "a large majority"? Is "generally" "a small majority"? Is "generally" "the largest minority"? Etc., etc.. You chose not to go to that level of precision because it was not necessary. And that's just for one sentence. Imagine an entire conversation with such provisos to consider.
Wait, wait, I think I see something here. I think I see why we are incapable of agreeing.
This seems more like a description of how S-O S's fail.
Can you offer any reason why I should treat S-O S's as a useful or realistic representational scheme if my goal is to draw accurate conclusions about actual, existing people?
Let me try to make my confusion clearer:
If I come upon a Halloween basket containing fifty peanut butter cups without razorblades, and ten peanut butter cups with razorblades, what is the second-order simulacrum I use to represent the contents of that basket? "A basket of delicious and safe peanut butter cups?"
Is this even a legitimate question, or am I still not grasping the concept?
There is a town. That town is called Simulacraton. Simulacraton is 40% white, 35% black, and 25% hispanic by population. The Joneses of Simulacraton -- are a semi-affluent suburban couple and live next door to a black man married to a hispanic woman. The Joneses are the second-order simulacrum of the average household in Simulacraton.
Second-order simulacra will always fail when you use them in ways that they are not meant to be used: such as actually being representative of individual instantiations of a thing: I.e.;, when you try to pretend they are anything other than an abstraction, a mapping of the territory designed for use as high-level overview to convey basic information without the need for great depth of inspection of the topic.
https://secure.wikimedia.org/wikipedia/en/wiki/Second-order_simulacra
The article says:
If I'm reading this correctly, it leaves me even more leery about the value of second-order simulacra.
Also from the article:
... did you intend for me to read this charitably? At best, it's a descriptive statement that says that people no longer care about the territory, and talk about maps without even realizing that they are not discussing territory. At worst, it says that reality has ceased to be real, which is Not Even Wrong.
If you want me to understand your ideas, please link me to clearer writing.
I am going to avoid using race or sex examples. I appreciate that you used Simulacraton as an object-level example, as it made your meaning much clearer, but I'd rather not discuss race when I am still unhappy with the resolution of the candy bowl problem.
I will revise my question for clarity:
"What is a reasonable second-order simulacrum of the contents of that basket of candy, and why? If no reasonable second-order simulacrum exists, why not?"
True, but none of the above reservations apply to the bowl of candy.
I am not claiming that the second-order simulacrum should represent the individual candies in the bowl. It may be wrong in any individual case. I am simply trying to convey a useful impression of the POPULATION, which is what you claim that SO S's are useful for.
I am not pretending that a simulacrum is anything more than an abstraction. I think it is a kind of abstraction that is not as useful as other kinds of abstraction when talking about populations.
I DO want a high-level overview, not a great depth of information. This overview should ideally reflect one REALLY important feature of the candy bowl.
(The statement that I would use to map the basket's population in detail would be "Ten of the sixty candies in the basket contain razorblades." The statement that I would use to map the basket broadly, without close inspection, would be, "Several of the candies in that basket contain razorblades."
if I had to use a second-order simulacrum, I would choose one of the candies with razorblades as my representative case, not the candy without. But this seems to break the plurality rule. Or perhaps, if feeling particularly perverse, I'd say "The candy in that basket contains one-sixth of a razorblade.")
I believe that second-order simulacra fail badly in the case of the candy basket. And if second-order simulacra can't handle simple hypothetical cases, shouldn't I be at least a little suspicious of this mapping strategy in general?
I'm hoping I can butt in and explain all this.
Logos01 probably shouldn't have brought up Baudrillard, who is among the sloppiest and most obscure thinkers of the last century. Baudrillard's model of abstraction is pretty terrible. Much better to user analytic philosophy's terminology rather than post-structuralism's terminology. In analytic philosophy we talk about abstract objects, "types" or "kinds". These are ubiquitous, not especially mysterious, and utterly essential to the representation of knowledge. "Electron", "Homo sapiens", "the combustion engine", "Mozart's 10th Symphony", "the Human Genome", etc. To map without abstract objects one would have to speak only of particulars and extensionally defined sets. And that's just the nouns-- whether one can even use verbs without recourse to abstraction is another issue entirely. Open up any scientific journal article and you will see named entities which are abstract objects. There are schools of thought that hold that kinds can ultimately be reduced to classes determined only by resemblance or predicate-- in an attempt to dissolve the supposed mystery of what abstract objects are. But even the most strident nominalists don't propose to actually do away with their usage.
None of that is particularly controversial and that's basically what Baudrillard means by "second-order simulacra". Now the question is, to what extent is it permissible to make statements about types which are not true of all of the particulars which instantiate that type? Call these "generalizations". We know from the limit cases that it can be both permissible and impermissible. "The Bobcat is found in North America" seems true and informative-- and yet there are bobcats in zoos outside that region. At the other end "Birds can talk" is mis-informative even though there are a few species of bird that can learn to talk.
The criterion for whether a generalization is permissible is chiefly pragamatic. You wouldn't say "The candy is safe" if there were a few razor blades mixed in because people are used to not having any razor blades mixed in at all! The fact about the candy that is worth communicating is that there are razor blades in a few of them. You're trying to warn people!
I think Logos's race examples above are wrong. Whether one specifies the race of the typical family depends on whether or not race is a relevant variable in what you are trying to communicate. If all you want to do is express to a Boston Red Sox fan that he or she shouldn't expect to find other fans in New York you would just say "New Yorkers don't like the Red Sox." There is no reason to say "New Yorkers are white people who don't like the Red Sox"-- even if the vast, vast majority of New Yorkers were white this would be communicating unnecessary detail given the goal of the communication. But if you're trying to constrain someone's expectations about what kind of people they will meet in New York saying "New Yorkers are white people who don't like the Red Sox" is mis-informative if most or a large fraction of New Yorkers aren't white people.
These are all simple examples which can be solved by adding another sentence at most. But in discussions of sufficient complexity additional specificity really does become untenable. At the limit demanding arbitrary precision would require you to use quantum field theory to build an airplane (Newtonian physics can be thought of as a generalization of quantum mechanics).
There are special cases. One is that people should include additional, irrelevant details in cases where not including them reinforces a popular belief that such details don't exist. This is especially true when the additional details are newly discovered. If one is speaking to a crowd that thinks, say, all men are heterosexual, it is worth qualifying statements about men that assume heterosexuality since not saying anything about the existence of homosexuals reinforces the false notion that they don't really exist or are extremely rare. When speaking to crowds who are very familiar with that information, qualifying it may look like additional, irrelevant information. Relatedly, when hearing about social identity no one likes to feel like they've been left out of the map. This is both an understandable feeling and an inevitable problem when trying to talk about issues involving social/cultural identity and experience. Almost always even the most carefully PC essay talking about how group x experience behavior y or institution z will ignore some subset of group x. Social types and kinds are particularly rife with exceptions-- there is simply too much individual variation. But at some point you have to generalize to talk about social identity. I think among respectful, tolerant and educated people it is helpful to maintain a constant policy of "Yes, we all know this isn't true for everyone but this is a useful generalization". Whether or not it makes sense for Less Wrong to adopt that policy is another question.
A statement like "Women want {thing}" leaves it unclear what the map is even supposed to be, barring clear context cues. This can lead to either fake disagreements or fake agreements.
Fake disagreements ("You said that Republicans are against gun control, but I know some who aren't!") are not too dangerous, I think. X makes the generalization, Y points out the exception, X says that it was a broad generalization, Y asks for more clarity in the future, X says Y was not being sufficiently charitable, and so on. Annoying to watch, but not likely to generate bad ideas.
Fake agreements can lead to deeper confusion. If X seriously believes that 99% of women have some property, and Y believes that only 80% of women have some property, then they may both agree with the generalization even if they have completely different ideas about what a charitable reading would be!
It costs next to nothing to say "With very few exceptions, women...", "A strong majority of women...." or "Most women...." The three statements mean different things, and establishing the meaning does not make communication next-to-impossible; it makes communication clearer. This isn't about charity, but clarity.
I in another subthread referenced the "Harry Potter and the Methods of Rationality" 'fanfic' written by Eliezer, when he mentioned how many fewer digits of Pi rational!Harry knew as compared to rational!Hermione.
The point is that I'm concerned not with charity nor with clarity, but rather with sufficiency to the current medium. Each of those little "costs next to nothing" statements actually do have a cost, one that isn't necessarily clear initially.
Are you familiar at all with how errors propagate in measurements? Each time you introduce new provisos, those statements affect the "informational value" of each dependent statement in its nest. This creates an analogous situation to the concept of significant digits in discourse.
For a topic like lukeprog's, in other words, the difference between 99% and 80% of women is below the threshold of significance. Eliminating it altogether (until such time as it becomes significant) is an important and valuable practice in communication.
Failure to effectively exercise that practice will result in needless 'clarifications' distracting from the intended message, hampering dialogs with unnecessary cognitive burden resultant from additional nesting of "informational quanta." In other words; if you add too many provisos to a statement, an otherwise meaningful and useful one will become trivially useless. An example of this in action can be found in another subthread of this conversation where someone stated he felt that there is a 'trend among frequent LessWrongers to over-generalize". This has informational meaning. He later added a 'clarification' that he hadn't intended the statement as an indication of population size, which totally reversed the informational value of his statement from an interesting one to a statement so utterly trivial that it is effectively without meaning or usefulness.
The cost of omitting them isn't clear initially, either.
I was generally taught to carry significant figures further than strictly necessary to avoid introducing rounding errors. If my final answer would have 3 significant digits, using a few buffer digits seemed wise. They're cheap.
Propagation of uncertainty is not a reason to drop qualifiers. It's a reason to use them. When reading an argument based on a generalization, I want to know the exceptions BEFORE the argument begins, not afterwards. That way, I can have a sense of how the uncertainties in each step affect the final conclusion.
If I want an answer to three significant figures, I do not begin my reasoning by rounding to two sigfigs, then trying to add in the last sigfig later.
If one person thinks that an argument depends on an assumption that fails in 1 in 100 cases, and someone else thinks the assumption fails in 1 in 5 cases, and they don't even know that they disagree, and pointing out this disagreement is regarded as some kind of map-territory error, they will have trouble even noticing when the disagreement has become significant.
This tends to happen to bad generalizations, yes. Once you consider all of the cases in which they are wrong, suddenly they seem to only be true in the trivial cases!
Good generalizations are still useful even after you have noted places where they are less likely to hold. Adding any number of true provisos will not make them trivial.
As for the cognitive load, why not state assumptions at the beginning of an essay where possible, rather than adding them to each individual statement? If the reader shares the assumptions, they'll just nod and move on. If the reader does NOT share the assumptions, then relieving them of the cognitive burden of being aware of disagreement is not a service.
I just now caught this, and... this is, I believe, where we have our fundamental disconnect.
By restricting the dialogue to essays the overwhelming majority of the meaningfulness of what I'm trying to say is entirely eliminated: my statements have been aimed at discussing the heuristic of measuring the cognitive burden per "unit" of information when communicating. The fact is that in a pre-planned document of basically any type one can safely assume a vastly greater available "pool of cognition" in his audience than in, say, a one-off comment in response to it, a youtube video comment, or something said over beers on a Friday night with your drinking-buddies.
I am struck by the thought that this metaphorically very similar to how Newton's classical mechanics equations manifest themselves from quantum mechanics after you introduce enough systems, or how the general relativity equations become effectively conventional at "non-relativistic" speeds: when you change the terms of the equations the apparent behaviors become significantly different. Just like how there's no need to bother considering your own relativistic mass when deciding whether or not to go on a diet, the heuristic I'm trying to discuss is vanishingly irrelevant to anything that one should expect from a thought-out-in advance, unrestricted-in-length, document.
For a moderately loose definition of 'thought out in advance', this describes most text-based, internet-based communication, and certainly the types of communication that can happen on LW.
I disagree with the usage of the term "moderately" here. I do not find it applicable. How many hours do you spend on each comment you make?
I don't see how your question is relevant to the topic at hand. I usually spend less than 15 minutes writing any given comment - most of mine are relatively short - but that's not counting time spent thinking about a topic and figuring out if I have something to say about it at all, which varies wildly and has been known to last days in some cases. But even in instances where I come up with a response near-instantly, it's generally because I've previously spent time thinking about the particular issue, and as a result have a high-quality cached response available, which certainly seems to fit the criteria for 'thought out in advance'!
Given that your personal commenting history on this site is extremely limited comparatively speaking I can't really say that I disagree with you directly on this.
But we weren't talking about just you personally, we were talking about "most text-based, internet-based communication". And you seem to be an exception, not a rule, when it comes to the normal dialogue/discourse I see in the commenting threads of LW. And LW itself is by far vastly the exception to the rule when it comes to dealing with statements made as a result from pre-formed thoughts.
That being said -- I would hope we can both agree that the notion that one can prepare for all possible conversations in advance regardless of topic is simply ludicrous without something resembling the heuristics I am trying to put a spotlight on.
Upvoted for clear communication.
I'm sort of puzzled, though, as to how I could have possibly interpreted your statements as applying to anything but the post and the comments on it; I saw no context clues suggesting that you meant "in everyday conversation." Did I miss these?
That said, if one of us had added just three or four words of proviso earlier, limiting our generalizations explicitly, we could have figured the disconnect out more quickly. I could have said that my generalizations apply best to essays and edited posts. You could have said that your generalizations apply best to situations where the added cost of qualifiers carries a higher burden.
Because we did not explicitly qualify our generalizations, but instead relied on context, we fell prey to a fake disagreement. However, any vindication I feel at seeing my point supported is nullified by the realization that I, personally, failed to apply the communication strategy that I was promoting.
Oops.
My language throughout was highly generalized. Consider my opening statement: "I am troubled by the vehemence by which people seem to reject the notion of using the language of the second-order simulacrum -- especially in communities that should be intimately aware of the concept that the map is not the territory."
And then also consider the fact that I used the term "discourse".
I didn't mean "everyday communication" specifically -- it simply is the venue where such a heuristic is most overtly valuable and noticeable. I did not qualify my generalizations because there were no qualifications to make: I was meaning the general sense.
Quite frankly, I did. That would be a modifying element to the "threshold of significance". (I.e.; "Is the cost of adding item X to this conversation greater than the value item X provides to the depth or breadth of information I am attempting to convey? If yes, do not add it. If no, do.") Because I was discussing so highly generalized a principle / heuristic, the fact that situations where added cost of qualifiers cost a higher burden is simply an inexorable conclusion from the assertion.
This seems like a context in which that shouldn't be expected to save you from unwarranted criticism and being misunderstood at all. ;-)
Well, it's tough: When I mean to be general and I use generalized terminology, should I not have the expectation of having communicated that my case is generalized?
Which is why I also discussed error propagation, which compounds.
I can only say that you are reading the metaphor too literally given the examples I've given thus far.
Of course!!! This isn't applicable to dialogue, however, as it has the opposite problem: the degree of cognitive burden to retain the informational value of a statement increases with the increased complexity. There is a limit on how much of this can be done in a given conversation.
Increasing complexity of statements to increase their accuracy can cause the ability to comprehend a statement to be reduced.
This statement carries a specific assumption of depth of dialogue which may or may not be valid.
Not adding those statements also has a cost.
Honestly, you don't know how many potential rationalists may find a post seemingly making unchallenged sweeping generalizations about women, and decide that these so-called rationalists are just a group of bigoted idiots that are less rational than your average person-in-the-street.
It's okay for someone to to say that pi is "3.14" if the other person knows that you know in reality it has more digits than that, and you're just being sufficient for your purposes. In short if there's actual transparency, not a double illusion of such.
But if they don't know that, if every post of yours may be perceived as an indication of complete positions (not hasty approximations thereof), it costs less to do things like say "most women" instead of "women" (or add a general disclaimer at the beginning) rather than not do it.
This is trivially true. What does adding them add to a conversation to which they are not relevant or significant?
This is uncontestably true. But the opposite is also true; you don't know how many potential rationalists may find a post filled with provisos and details and, upon discovering a massive gulf of an inferential gap, give up on even attempting to understand.
Certainly.
This is a gross misrepresentation of my statements, to the point of being nothing remotely like what I advocate. I have repeatedly advocated not the elimination of precision but the application of only the relevant degree of precision to the nature of the discourse at hand.
My point is not restricted to '''"most women" instead of "women"'''. It is a generalized principle which happens to apply here. For any given conversation there are thousands of such details we must choose to parse for relevance to a conversation. Demanding unerring accuracy beyond relevance is simply damaging to dialogue.
And yet, we still say that p(Christianity is correct) is epsilon, rather than zero - and this seems to cause few-to-no problems, even.
Seems is the key here. Any instance where you would use that sort of language, the relevant threshold of significance was such that it was a proper statement to make.
Consider a context where you were making that statement to a Jehovah's Witness trying to hand you a flyer as your 10 o'clock bus was stopping in front of you. You could still make the statement, but if you were being honest with yourself you'd realize that your words would be gibberish, whereas "I'm not Christian" would be contextually appropriate: you would convey a statement with non-zero informational value. "The probability that 'Christianity is correct' is epsilon" on the other hand would not in such a context, quite likely, actually convey any meaning to the audience.
It seems that I've failed to make my point.
It is, as far as I can tell, safe to assume that everyone who reads LW understands enough about probabilities that saying 'zero' would communicate exactly the same concept regarding the probability at hand as saying 'epsilon', if we had a norm of allowing the former. The reason for doing the latter is about signaling, in much the same way that saying 'most women' instead of just 'women' is about signaling. In both cases, the point of the signal is to encourage accurate thought in the long run, rather than letting a small amount of convenience in the near term to outweigh that.
Either you have or I have. As I believe I understand entirely what your position here is, I can't help but wonder.
Here's the thing: nothing I've been saying was tailored at any point to be specific to Less Wrong in particular.
It's also not a safe assumption, by the way, for the simple fact there is at least one person who recommends this community to every budding (or potential) rationalist he encounters -- me. At least one of those persons (my ex-primary of 10 years) has an exceedingly poor capability of grasping mathematics and probabilities. This was one of the reasons she and I didn't make it past that 10 year mark.
See, I suspect there might be a political element to this as well. I for one would strongly prefer that the second-order simulacrum be the standard assumption rather than requiring continued increased cognitive burden in discourse. It is true that we think in language; and therefore the language we use shapes our thoughts -- but language is a memeplex of symbolic representations of semantical content/value. If we adjust the symbol, we adjust the thought. But this is now becoming an altogether different topic of conversation.
Reductively, the long term is nothing more than a collection of near terms. What remains a constant near term burden over the long term becomes a long-term burden.
I remain of the position that constantly adding caveats and provisos to language regardless of where the focus of discourse at a given moment happens to be is a fundamental error in communication. Since we can't seem to agree on this topic, I have to wonder what postulates we aren't sharing in common.
Not judging but... this is a very novel reason for ending a 10-year relationship.
"One of" is a key term here. I also didn't provide any context for weighting of said reasons.
I didn't make those clarifications because it really wasn't relevant to the information I was trying to convey at the time. ;-)