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.
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.
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.
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 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.
Not adding those statements also has a cost.
in other words, the difference between 99% and 80% of women is below the threshold of significance.
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?
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.
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.
[Re: Pi “is” 3.14] In short if there’s actual transparency, not a double illusion of such.
Certainly.
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)
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.
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.
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.
Each of those little “costs next to nothing” statements actually do have a cost, one that isn’t necessarily clear initially.
The cost of omitting them isn’t clear initially, either.
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.
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.
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.
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.
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.
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.
As for the cognitive load, why not state assumptions at the beginning of an essay where possible,
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.
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.
I saw no context clues suggesting that you meant “in everyday conversation.” Did I miss these?
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.
You could have said that your generalizations apply best to situations where the added cost of qualifiers carries a higher burden.
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.
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?
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 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’!
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,
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.
How many hours do you [emphasis yours] spend on each comment you make?
If you’re going to change the subject, at least don’t try to act like I’m doing something wrong when I politely go along with the subject change, okay?
we were talking about “most text-based, internet-based communication”.
Most text-based, internet-based communication has very little in the way of time pressure, and LessWrong specifically has a norm of allowing or even encouraging comments on older posts and comments, allowing for arbitrary levels of pre-thinking. Length restrictions are slightly more common on the internet at large, but still not the norm, and not present here. This, in the context of your original comment—plus the implication that since it is possible to do those things, any case where someone doesn’t is a matter of personal choice or (problematic, in my opinion) group norms—was the entirety of my original point.
I do agree that the idea of having cached responses to all conversational possibilities is ridiculous. I wasn’t proposing that that is a thing that people should particularly try to do. My point, insofar as I had a point and wasn’t just answering your question on the assumption that you had some use for the information, was that that is one of the tactics that I’ve found to work, the other main one being to actually take the time to think my responses through, even if that takes a while.
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.
I am not at all sure what you’re trying to communicate, here. One possible way of parsing it suggests that you might think that since LW is already well above average in terms of good communication, making it better shouldn’t be a priority, which I disagree with. I’d strongly prefer a clarification of your actual intent to a discussion of that idea if it wasn’t what you were trying to communicate, though.
How many hours do you [emphasis yours] spend on each comment you make?
If you’re going to change the subject, at least don’t try to act like I’m doing something wrong when I politely go along with the subject change, okay?
I was using an example to demonstrate the intended meaning (which apparently was not a well-aimed one given the fact that you are statistically aberrant). I was not changing the topic.
Most text-based, internet-based communication has very little in the way of time pressure,
If I cared about time pressure as opposed to cognitive burden -- that is, available attention span—I would have indicated so. I don’t, so this isn’t relevant.
and LessWrong specifically has a norm of allowing or even encouraging comments on older posts and comments, allowing for arbitrary levels of pre-thinking.
Even so, my point remains easily demonstrated by a perusal of the majority of comments, which are typically made in a “conversational” rather than “ex post facto” mode. (We, right now, are in that conversational mode.)
This, in the context of your original comment—plus the implication that since it is possible to do those things, so any cases where someone doesn’t is a matter of personal choice or (problematic, in my opinion) group norms—was the entirety of my original point.
A) that wasn’t my original comment.
B) Your counter-point as I understand it still remains invalid, to be quite honest, because you’re—I cannot help but feel intentionally at this point—refusing to recognize the fact that you’re using statistical outliers instead of norms to support your claims against what I have already stated explicitly was a heuristic.
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.
I am not at all sure what you’re trying to communicate, here. One possible way of parsing it suggests that you might think that since LW is already well above average in terms of good communication, making it better shouldn’t be a priority,
No, that is not a valid interpretation of my statement. You leave out the context provided by antecedent statement of mine (same comment) that necessarily influences the meaning: “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.” It is clear that how I said you were different was in that you have a limited commenting history.
I’d strongly prefer a clarification of your actual intent to a discussion of that idea if it wasn’t what you were trying to communicate, though.
I seem to have some strong difficulties in communicating with you any of my intended meanings at pretty much any point. I’m not at all certain why this is the case, as I do not normally have this difficulty with an audience. I have noted that you have left out contextually significantly relevant points/items in coming to your interpretations of my words as I have written them.
I do not know why that is happening, but it makes me feel that this conversation is never going to go anywhere but frustrate me. So no, you won’t get that clarification; but not because I wouldn’t like to give it.
I was generally taught to carry significant figures further than strictly necessary to avoid introducing rounding errors.
Which is why I also discussed error propagation, which compounds.
Propagation of uncertainty is not a reason to drop qualifiers. It’s a reason to use them.
I can only say that you are reading the metaphor too literally given the examples I’ve given thus far.
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.
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.
If the reader does NOT share the assumptions, then relieving them of the cognitive burden of being aware of disagreement is not a service.
This statement carries a specific assumption of depth of dialogue which may or may not be valid.
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 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.
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.
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.
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.
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.
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.
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.
(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.
Not judging but… this is a very novel reason for ending a 10-year relationship.
However, the simple truth is that communication becomes positively impossible if ‘sweeping generalizations’ at some > level are not made.
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.
True but misleading. One should seek to avoid eliminating relevant meaning in the process of making those generalizations.
(Formatting tip: you need to add two spaces at the end of the previous line to get lesswrong’s commenting markup language to “ ”/”\n”. Two newlines will ”
″.)
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.
When someone adds that proviso “asexual/homosexual”—they are changing the relevant level of precision necessary to the conversation.
No, they are pointing out that in order to apply to a case they are interested in, the conversation must be made more precise.
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.
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.
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 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.)
When someone adds that proviso “asexual/homosexual”—they are changing the relevant level of precision necessary to the conversation.
No, they are pointing out that in order to apply to a case they are interested in, the conversation must be made more precise.
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 last one isn’t a distraction, it’s a counterexample.
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.
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.
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].)
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”.
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.
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.
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.
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.
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.
Is this even a legitimate question, or am I still not grasping the concept?
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.
Second-order simulacra, a term coined by Jean Baudrillard, are symbols without referents, that is, symbols with no real object to represent. Simply put, a symbol is itself taken for reality and further layer of symbolism is added. This occurs when the symbol is taken to be more important or authoritative of the original entity, authenticity has been replaced by copy (thus reality is replaced by a substitute).
If I’m reading this correctly, it leaves me even more leery about the value of second-order simulacra.
Also from the article:
Baudrillard argues that in the postmodern epoch, the territory ceases to exist, and there is nothing left but the map; or indeed, the very concepts of the map and the territory have become indistinguishable, the distinction which once existed between them having been erased.
… 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?”
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.
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?
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.
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.
Using ‘x’, ‘y’, and ‘z’ as labels to represent variable groups reinforces the pernicious stereotype that other letters aren’t worthy of being used as labels to represent variables and don’t count.
I don’t appreciate how lazy these jokes are. Once posting on LW one would assume unnecessary tribal signaling in the form of easy, form-fillable potshots at the religious, “political correctness,” non-nerd popular culture, &c.
After I write a six-paragraph explanation of abstraction and the pragmatics of generalization I reserve the right to tell a lazy joke.
I think you’re reading too much into the joke though. I wasn’t intending to make fun of political correctness- hopefully what I wrote before makes it clear that that is not my attitude. I did find lessdazed comment humorous both for the meta-ness of turning the subject of the paragraph back on the text itself and for the juxtaposition of the concern for inclusiveness being applied to silly, non-human things like variable letters. So I played along. The joke was a good way of emphasizing that that particular concern about generalizations is not about communication or accuracy, but about how we treat people.
Whether lessdazed was trying to make fun of political correctness or not you’ll have to ask him.
I habsolutely zero intentions. I had hoped that you would be capable of being a rational agent in this dialogue. If, however, that isn’t something you care to do, we can end this conversation here and now.
“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?”
The stereotypical bowl-candy is perfectly safe. It likely has a neighbor that has a razorblade in it.
-- In a side-note, why did you feel the need to push this particular variation of your question on me when I had already answered it? What, exactly, did you think the Simulacraton example was? Or did you not make the connection merely because you used candies and ratios and I used people and percents?
I believe that second-order simulacra fail badly in the case of the candy basket.
Of COURSE they do. It’s not an applicable or relevant scenario in which one SHOULD use a second-order simulacrum in.
The scale is vastly too small to allow for abstraction to be useful.
The topic at hand focuses on the group in question rather than some other topic to which the group is tangential.
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?
Good luck getting through life without ever constructing a symbolic representation of anything at any time ever under any circumstances: because that’s what you are arguing against.
Downvoted for telling me what I’m arguing for and against, for something like the third time now, when I am fairly certain that our intuitive ideas of how abstraction works are somewhat different. This is one of the few things that breaks my internal set of “rules for a fair argument.”t.
(Note: I am NOT downvoting for the paragraph beginning “OF COURSE they do”, because it’s given me a hunch as to what is going on here, is clearly written, and makes your actual objections to the candy bowl case clearer.
I SHOULD not be downvoting for the first paragraph, but it affected the decision.)
I habsolutely zero intentions. I had hoped that you would be capable of being a rational agent in this dialogue. If, however, that isn’t something you care to do, we can end this conversation here and now.
When I tried to work out what you meant by second-order simulacra, you linked me to a cryptic Wikipedia article discussing a vague description of the term, along with confused-looking statements about the nature of reality. I really did NOT know what your intentions were, and I genuinely was getting exasperated.
I am sorry for implying bad faith. I should have said, “I have no clue what I am supposed to take from this article, but it sends extremely dubious signals to me about the validity of this concept.”
In a side-note, why did you feel the need to push this particular variation of your question on me when I had already answered it?
Because you hadn’t. I presented an example where second-order simulacra fail. Reading the reply, I was unsatisfied to find a description of a different case, followed by a statement that second-order simulacra fail in the candy bowl case, but for reasons that weren’t consistent with the example.
What, exactly, did you think the Simulacraton example was?
An example chosen in which your heuristic gave a semi-plausible answer, when I had asked about a place where it ceases to work.
Or did you not make the connection merely because you used candies and ratios and I used people and percents?
I did. I did not conceive, however, that your answer would be:
The stereotypical bowl-candy is perfectly safe. It likely has a neighbor that has a razorblade in it.
The analogy to the population of people was stretched enough—and not just for reasons of ratios and percents—that there was no WAY I’d come to the above answer without questioning it.
The scale is vastly too small to allow for abstraction to be useful.
The topic at hand focuses on the group in question rather than some other topic to which the group is tangential.
This is getting closer to what I actually am looking for—a situation where I ought to use second-order simulacra. However, I still do not think these are problems for the candy bowl.
1: Abstractions can work on an arbitrarily small sample size. “A bowl of candies, some of which are unsafe” IS an abstraction.. If that is not abstract enough, what about a pie chart showing the proportion of unsafe candies?
If a group is truly tangential to a topic, how do you decide which features are important enough to include in your abstraction? Why include ANY features in your abstraction besides “lives in Simulacraton?” It does no good to say that one would abstract the Joneses as being of the plurality race. For example, I could imagine them as being racially indeterminate. But I have trouble imagining them at all.
Good luck getting through life without ever constructing a symbolic representation of anything at any time ever under any circumstances: because that’s what you are arguing against.
Generally speaking, that is not representations work in my mind. The phrase “generalizing from one example” is ringing a bell right now.
When I am told “the population of Simulacraton is 40% white,” I don’t really feel any need to abstractly represent the population with one person, neighbors or no, or to refer to such a person in conversation. I would not say, “People from Simulacraton are {X},” and I tend to react to such statements with skepticism because I see them as unqualified statements about an entire set of people based on weak evidence.
How do I describe the average family in that town? With reluctance. I default to mapping by groups. In fact, I’m not used to visual or instance-based representation in general. It may be developmental—I was born blind and raised blind for a month before surgery. This may have affected my brain development in odd ways; I’m still bad with faces.
It does seem likely to me that a more visual thinker would find it convenient to imagine an average family as having visibly defined properties representing a plurality, rather than properties that can’t be visually imagined as easily. But my ‘average member’ is just a bunch of loosely defined properties tied together with a name, and many of the properties that are needed to visualize a person clearly are missing from that set.
I don’t think ONLY in verbally described sets, of course. I also think in free-floating sensory memories that rarely remain in my consciousness for very long. But “thinks in sets defined by verbal descriptions” is a good approximation of what I do.
Example: I have never been to Paris. If I were to talk about the Eiffel Tower, and for some reason felt the need to mention a Parisian in the description, I would likely say “a Parisian.” I wouldn’t give them a name or any properties unless I had to. If I did, the properties would be based on what I saw in movies, not any properties that reflect a plurality of Parisians, and I would assign them in a miserly way. My second-order simulacrum would be useless for anything but fake local flavor.
What about questions where “a Parisian” is just a tangential feature, where precision in the description of the Parisian is unimportant? Surely I use a second-order simulacrum then, right?
Nope.
For me, it is cognitively cheaper to not reference “a typical Parisian” when asked a question that tangentially involves people from Paris, because that would require me to represent a typical Parisian symbolically, and I have trouble imagining such a thing as “a typical Parisian.” Instead, I would simply say, “a random Parisian,” and my mental representation of such a Parisian would be the word “Parisian” with attached possible properties, half-formed images, and phrases spoken in movies.
THIS is why qualifiers like “almost always,” “generally”, “about half of the time,” “on occasion”, and “almost never” strike me as informative—they are quick and dirty ways to adjust the sets in my head! They are cognitively cheap for me, though not NEARLY as cheap for me as numerical probability estimates, which are great when people actually bother to give them.
Now, I am not naive enough to think that a “set” is part of the territory itself, but once one starts to cluster entities together, using a second-order generalization may reinforce confusion about the properties of entities in that cluster. When I discourage the use of second-order simulacra without disclaimers, it is not because I fail to realize my set-based map is not the territory, but because many people will name a cluster of entities, pick a single entity from that cluster, generalize to the entire cluster, and imagine that they have actually described a lot of territory in a useful way.
People do this constantly in politicized arguments. Context is not enough, and the more unwilling someone is to add a proviso, the more I suspicious I grow of the reasons that they are unwilling to do so. I suspect that my attitude towards unqualified generalizations is very similar to your attitude toward qualified generalizations. They seem like useless maps to me because I don’t use them and don’t really know how to.
Context is not enough, and the more unwilling someone is to add a proviso, the more I suspicious I grow of the reasons that they are unwilling to do so.
Is this a matter of degree or of kind? It seems to me like the issue here is how many qualifications should be made in particular contexts, and so is a question of degree, and not at all one of kind. This means that there is a possible mind with standards analogous to yours to the same degree yours are analogous to Logos01.
For example, where Logos01 thinks an essay with five paragraphs of content needs one disclaimer, you might think it needs fifty, and some third party might think it needs two thousand and fifty, and some fourth party 125,000. Any criticism you apply to him or her seems applicable to you as well, for all trade off precision for brevity.
It therefore seems impossible to muster a strong argument against Logos01′s general practice of being imprecise for the sake of finishing sentences despite lack of perfect precision, because you do that as well, and so it seems your argument can’t be stronger than a weaker one against a particular balance of trade offs.
Downvoted for telling me what I’m arguing for and against, for something like the third time now, when I am fairly certain that our intuitive ideas of how abstraction works are somewhat different. This is one of the few things that breaks my internal set of “rules for a fair argument.”t.
I made no “intuitive” statements about “how abstraction works”. Ever.
Your positional statements made it quite clear that your objection to S-O S’s was in the fact that they are an abstraction.
You repeatedly made several arbitrary statements about representative symbols and how they would “have” to be that I demonstrated to be inaccurate of how abstraction is done.
I never make the statement, “You are arguing X” unless it is factually and demonstrably true. You stated that you “distrusted” “this method” (“this method” being the use of symbols without referents) of abstraction… but unfortunately, that’s all abstraction is; “making maps.” If you don’t like it when someone tells you what you are or aren’t arguing for or against, don’t put yourself into a position where those statements would be true. If you had said, “The sky is blue”, and I told you, “You are saying the sky is blue”, would you also react so childishly?
The rest of your post is simply too long for me to bother with. This topic has gone beyond my threshold of conversational utility: you demonstrate that you will accept nothing I say at any point and are merely arguing for the sake of arguing. Case in point:
If a group is truly tangential to a topic, how do you decide which features are important enough to include in your abstraction?
They are topical. This is a tautology. And this marks at least the second time I’ve called out your continuing to riddle the topic with questions that have already been answered or have answers whose very questions demonstrate them. This is not the mark of an honest conversant.
Further:
Generally speaking, that is not representations work in my mind. The phrase “generalizing from one example” is ringing a bell right now.
This directly contradicts the very definition of the word, “abstraction”. Abstraction—and mental representation is never anything BUT abstraction—is definitionally constructing simulacra within the mind.
I point this out as yet another demonstrative example of your arguing for what I can only describe merely the sake of arguing.
Rounding this out:
I suspect that my attitude towards unqualified generalizations is very similar to your attitude toward qualified generalizations. They seem like useless maps to me because I don’t use them and don’t really know how to.
No. This is a flat-out false characterization of my position and I have explicitly disagreed with it. I said nothing of the sort. Ever. And I haven’t been arguing in favor of such a position.
You are A) misrepresenting me. B) refusing to accept basic definitions of terminology relevant to this topic, C) continuously raising questions that have already been answered, amongst other things.
I’ll not be responding to you in this topic again.
For example; if I say “Men and women get married because they love each other”,
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.
Oooh, perfect example! Because this is probably still not true for a plurality, if not majority of humanity,
[...]
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.
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.
the assumption that the listeners fit their relatively vague preconception of who they talk to, rather than about.
Can you rephrase this for me? It’s not parsing my language-interpreter.
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.
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.
an English-speaking, technically-trained man in India (where arranged marriages for purposes other than romantic love are still pretty standard).
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.
Sometimes people on this site even take norms like that and try to infer over all of human evolution.
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.
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.”
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.
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.
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.
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.
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?
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.
o.O
If you’re going to change the subject, at least don’t try to act like I’m doing something wrong when I politely go along with the subject change, okay?
Most text-based, internet-based communication has very little in the way of time pressure, and LessWrong specifically has a norm of allowing or even encouraging comments on older posts and comments, allowing for arbitrary levels of pre-thinking. Length restrictions are slightly more common on the internet at large, but still not the norm, and not present here. This, in the context of your original comment—plus the implication that since it is possible to do those things, any case where someone doesn’t is a matter of personal choice or (problematic, in my opinion) group norms—was the entirety of my original point.
I do agree that the idea of having cached responses to all conversational possibilities is ridiculous. I wasn’t proposing that that is a thing that people should particularly try to do. My point, insofar as I had a point and wasn’t just answering your question on the assumption that you had some use for the information, was that that is one of the tactics that I’ve found to work, the other main one being to actually take the time to think my responses through, even if that takes a while.
I am not at all sure what you’re trying to communicate, here. One possible way of parsing it suggests that you might think that since LW is already well above average in terms of good communication, making it better shouldn’t be a priority, which I disagree with. I’d strongly prefer a clarification of your actual intent to a discussion of that idea if it wasn’t what you were trying to communicate, though.
I was using an example to demonstrate the intended meaning (which apparently was not a well-aimed one given the fact that you are statistically aberrant). I was not changing the topic.
If I cared about time pressure as opposed to cognitive burden -- that is, available attention span—I would have indicated so. I don’t, so this isn’t relevant.
Even so, my point remains easily demonstrated by a perusal of the majority of comments, which are typically made in a “conversational” rather than “ex post facto” mode. (We, right now, are in that conversational mode.)
A) that wasn’t my original comment.
B) Your counter-point as I understand it still remains invalid, to be quite honest, because you’re—I cannot help but feel intentionally at this point—refusing to recognize the fact that you’re using statistical outliers instead of norms to support your claims against what I have already stated explicitly was a heuristic.
No, that is not a valid interpretation of my statement. You leave out the context provided by antecedent statement of mine (same comment) that necessarily influences the meaning: “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.” It is clear that how I said you were different was in that you have a limited commenting history.
I seem to have some strong difficulties in communicating with you any of my intended meanings at pretty much any point. I’m not at all certain why this is the case, as I do not normally have this difficulty with an audience. I have noted that you have left out contextually significantly relevant points/items in coming to your interpretations of my words as I have written them.
I do not know why that is happening, but it makes me feel that this conversation is never going to go anywhere but frustrate me. So no, you won’t get that clarification; but not because I wouldn’t like to give it.
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.
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. ;-)
Also, a factor like that may have been a significant cause of other more proximate issues.
Okay, I’ll admit it—that just got a grin and a chuckle out of me. Well done.
bows
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 “
”/”\n”. Two newlines will ”
″.)
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.
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.
Using ‘x’, ‘y’, and ‘z’ as labels to represent variable groups reinforces the pernicious stereotype that other letters aren’t worthy of being used as labels to represent variables and don’t count.
I don’t appreciate your attempt to erase the experiences of the Greek alphabet!
I don’t appreciate how lazy these jokes are. Once posting on LW one would assume unnecessary tribal signaling in the form of easy, form-fillable potshots at the religious, “political correctness,” non-nerd popular culture, &c.
After I write a six-paragraph explanation of abstraction and the pragmatics of generalization I reserve the right to tell a lazy joke.
I think you’re reading too much into the joke though. I wasn’t intending to make fun of political correctness- hopefully what I wrote before makes it clear that that is not my attitude. I did find lessdazed comment humorous both for the meta-ness of turning the subject of the paragraph back on the text itself and for the juxtaposition of the concern for inclusiveness being applied to silly, non-human things like variable letters. So I played along. The joke was a good way of emphasizing that that particular concern about generalizations is not about communication or accuracy, but about how we treat people.
Whether lessdazed was trying to make fun of political correctness or not you’ll have to ask him.
I habsolutely zero intentions. I had hoped that you would be capable of being a rational agent in this dialogue. If, however, that isn’t something you care to do, we can end this conversation here and now.
The stereotypical bowl-candy is perfectly safe. It likely has a neighbor that has a razorblade in it.
-- In a side-note, why did you feel the need to push this particular variation of your question on me when I had already answered it? What, exactly, did you think the Simulacraton example was? Or did you not make the connection merely because you used candies and ratios and I used people and percents?
Of COURSE they do. It’s not an applicable or relevant scenario in which one SHOULD use a second-order simulacrum in.
The scale is vastly too small to allow for abstraction to be useful.
The topic at hand focuses on the group in question rather than some other topic to which the group is tangential.
Good luck getting through life without ever constructing a symbolic representation of anything at any time ever under any circumstances: because that’s what you are arguing against.
Downvoted for telling me what I’m arguing for and against, for something like the third time now, when I am fairly certain that our intuitive ideas of how abstraction works are somewhat different. This is one of the few things that breaks my internal set of “rules for a fair argument.”t.
(Note: I am NOT downvoting for the paragraph beginning “OF COURSE they do”, because it’s given me a hunch as to what is going on here, is clearly written, and makes your actual objections to the candy bowl case clearer.
I SHOULD not be downvoting for the first paragraph, but it affected the decision.)
When I tried to work out what you meant by second-order simulacra, you linked me to a cryptic Wikipedia article discussing a vague description of the term, along with confused-looking statements about the nature of reality. I really did NOT know what your intentions were, and I genuinely was getting exasperated.
I am sorry for implying bad faith. I should have said, “I have no clue what I am supposed to take from this article, but it sends extremely dubious signals to me about the validity of this concept.”
Because you hadn’t. I presented an example where second-order simulacra fail. Reading the reply, I was unsatisfied to find a description of a different case, followed by a statement that second-order simulacra fail in the candy bowl case, but for reasons that weren’t consistent with the example.
An example chosen in which your heuristic gave a semi-plausible answer, when I had asked about a place where it ceases to work.
I did. I did not conceive, however, that your answer would be:
The analogy to the population of people was stretched enough—and not just for reasons of ratios and percents—that there was no WAY I’d come to the above answer without questioning it.
This is getting closer to what I actually am looking for—a situation where I ought to use second-order simulacra. However, I still do not think these are problems for the candy bowl.
1: Abstractions can work on an arbitrarily small sample size. “A bowl of candies, some of which are unsafe” IS an abstraction.. If that is not abstract enough, what about a pie chart showing the proportion of unsafe candies?
If a group is truly tangential to a topic, how do you decide which features are important enough to include in your abstraction? Why include ANY features in your abstraction besides “lives in Simulacraton?” It does no good to say that one would abstract the Joneses as being of the plurality race. For example, I could imagine them as being racially indeterminate. But I have trouble imagining them at all.
Generally speaking, that is not representations work in my mind. The phrase “generalizing from one example” is ringing a bell right now.
When I am told “the population of Simulacraton is 40% white,” I don’t really feel any need to abstractly represent the population with one person, neighbors or no, or to refer to such a person in conversation. I would not say, “People from Simulacraton are {X},” and I tend to react to such statements with skepticism because I see them as unqualified statements about an entire set of people based on weak evidence.
How do I describe the average family in that town? With reluctance. I default to mapping by groups. In fact, I’m not used to visual or instance-based representation in general. It may be developmental—I was born blind and raised blind for a month before surgery. This may have affected my brain development in odd ways; I’m still bad with faces.
It does seem likely to me that a more visual thinker would find it convenient to imagine an average family as having visibly defined properties representing a plurality, rather than properties that can’t be visually imagined as easily. But my ‘average member’ is just a bunch of loosely defined properties tied together with a name, and many of the properties that are needed to visualize a person clearly are missing from that set.
I don’t think ONLY in verbally described sets, of course. I also think in free-floating sensory memories that rarely remain in my consciousness for very long. But “thinks in sets defined by verbal descriptions” is a good approximation of what I do.
Example: I have never been to Paris. If I were to talk about the Eiffel Tower, and for some reason felt the need to mention a Parisian in the description, I would likely say “a Parisian.” I wouldn’t give them a name or any properties unless I had to. If I did, the properties would be based on what I saw in movies, not any properties that reflect a plurality of Parisians, and I would assign them in a miserly way. My second-order simulacrum would be useless for anything but fake local flavor.
What about questions where “a Parisian” is just a tangential feature, where precision in the description of the Parisian is unimportant? Surely I use a second-order simulacrum then, right?
Nope.
For me, it is cognitively cheaper to not reference “a typical Parisian” when asked a question that tangentially involves people from Paris, because that would require me to represent a typical Parisian symbolically, and I have trouble imagining such a thing as “a typical Parisian.” Instead, I would simply say, “a random Parisian,” and my mental representation of such a Parisian would be the word “Parisian” with attached possible properties, half-formed images, and phrases spoken in movies.
THIS is why qualifiers like “almost always,” “generally”, “about half of the time,” “on occasion”, and “almost never” strike me as informative—they are quick and dirty ways to adjust the sets in my head! They are cognitively cheap for me, though not NEARLY as cheap for me as numerical probability estimates, which are great when people actually bother to give them.
Now, I am not naive enough to think that a “set” is part of the territory itself, but once one starts to cluster entities together, using a second-order generalization may reinforce confusion about the properties of entities in that cluster. When I discourage the use of second-order simulacra without disclaimers, it is not because I fail to realize my set-based map is not the territory, but because many people will name a cluster of entities, pick a single entity from that cluster, generalize to the entire cluster, and imagine that they have actually described a lot of territory in a useful way.
People do this constantly in politicized arguments. Context is not enough, and the more unwilling someone is to add a proviso, the more I suspicious I grow of the reasons that they are unwilling to do so. I suspect that my attitude towards unqualified generalizations is very similar to your attitude toward qualified generalizations. They seem like useless maps to me because I don’t use them and don’t really know how to.
Is this a matter of degree or of kind? It seems to me like the issue here is how many qualifications should be made in particular contexts, and so is a question of degree, and not at all one of kind. This means that there is a possible mind with standards analogous to yours to the same degree yours are analogous to Logos01.
For example, where Logos01 thinks an essay with five paragraphs of content needs one disclaimer, you might think it needs fifty, and some third party might think it needs two thousand and fifty, and some fourth party 125,000. Any criticism you apply to him or her seems applicable to you as well, for all trade off precision for brevity.
It therefore seems impossible to muster a strong argument against Logos01′s general practice of being imprecise for the sake of finishing sentences despite lack of perfect precision, because you do that as well, and so it seems your argument can’t be stronger than a weaker one against a particular balance of trade offs.
I made no “intuitive” statements about “how abstraction works”. Ever.
Your positional statements made it quite clear that your objection to S-O S’s was in the fact that they are an abstraction.
You repeatedly made several arbitrary statements about representative symbols and how they would “have” to be that I demonstrated to be inaccurate of how abstraction is done.
I never make the statement, “You are arguing X” unless it is factually and demonstrably true. You stated that you “distrusted” “this method” (“this method” being the use of symbols without referents) of abstraction… but unfortunately, that’s all abstraction is; “making maps.” If you don’t like it when someone tells you what you are or aren’t arguing for or against, don’t put yourself into a position where those statements would be true. If you had said, “The sky is blue”, and I told you, “You are saying the sky is blue”, would you also react so childishly?
The rest of your post is simply too long for me to bother with. This topic has gone beyond my threshold of conversational utility: you demonstrate that you will accept nothing I say at any point and are merely arguing for the sake of arguing. Case in point:
They are topical. This is a tautology. And this marks at least the second time I’ve called out your continuing to riddle the topic with questions that have already been answered or have answers whose very questions demonstrate them. This is not the mark of an honest conversant.
Further:
This directly contradicts the very definition of the word, “abstraction”. Abstraction—and mental representation is never anything BUT abstraction—is definitionally constructing simulacra within the mind.
I point this out as yet another demonstrative example of your arguing for what I can only describe merely the sake of arguing.
Rounding this out:
No. This is a flat-out false characterization of my position and I have explicitly disagreed with it. I said nothing of the sort. Ever. And I haven’t been arguing in favor of such a position.
You are A) misrepresenting me. B) refusing to accept basic definitions of terminology relevant to this topic, C) continuously raising questions that have already been answered, amongst other things.
I’ll not be responding to you in this topic again.
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.
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.