The value of ambiguous speech
This was going to be a reply in a discussion between ChristianKl and MattG in another thread about conlangs, but their discussion seemed to have enough significance, independent of the original topic, to deserve a thread of its own. If I’m doing this correctly (this sentence is an after-the-fact update), then you should be able to link to the original comments that inspired this thread here: http://lesswrong.com/r/discussion/lw/n0h/linguistic_mechanisms_for_less_wrong_cognition/cxb2
Is a lack of ambiguity necessary for clear thinking? Are there times when it’s better to be ambiguous? This came up in the context of the extent to which a conlang should discourage ambiguity, as a means of encouraging cognitive correctness by its users. It seems to me that something is being taken for granted here, that ambiguity is necessarily an impediment to clear thinking. And I certainly agree that it can be. But if detail or specificity are the opposites of ambiguity, then surely maximal detail or specificity is undesirable when the extra information isn’t relevant, so that a conlang would benefit from not requiring users to minimize ambiguity.
Moving away from the concept of conlangs, this opens up some interesting (at least to me) questions. Exactly what does “ambiguity” mean? Is there, for each speech act, an optimal level of ambiguity, and how much can be gained by achieving it? Are there reasons why a certain, minimal degree of ambiguity might be desirable beyond avoiding irrelevant information?
Sometimes you don’t even notice that your language has an ambiguity problem until you try translating it.
In my native Spanish, “to wish” and “to desire” are expressed by the same word, and it took me years to realize that the English “wish” and “desire” were different concepts. (This was a huge source of confusion when I was studying Buddhism.)
Similarly, “to await”, “to hope” and “to expect” are the same word in Spanish. (One song in Spanish by Shakira has a funny verse that uses the same word three times, but by the specific usage you can tell it should translate as “I hope you don’t expect me to wait for you.”)
I don’t know what specific circumstances prompted Spanish to conflate those sets of concepts under unifying words (or maybe what happened was that English had some reason to separate originally closer concepts; I don’t know), but this is a big blind spot for people not exposed to other languages.
English has a big ambiguity problem with the word “love”...
English does have words like crave, cherish, yearn and lust to specify certain subtypes of love. It’s just that people don’t like to use them and prefer to use a word that’s more ambigious. I don’t think it’s a problem that English offers the ability to be ambigious in this instance.
Precise language is great when you have a precise message. Often what we are trying to convey is itself not precise.
If words paint mental pictures, then precise language is like sending a photograph where everything is in full focus. It’s great in that there’s a lot of information there, but it’s not always clear what to do with it and focusing on the wrong bits can get in the way of an important message.
Instead you can turn down the depth of field such that the only thing in any focus is the object you’re pointing at—and even then only in sufficient detail to identify it and no more. Then you have no choice but to recognize the actual intended message because you were careful not to dilute your point with extraneous information that often comes with careless use of language.
If I’m pointing out a tiger, not only do I want to make sure you don’t get distracted looking at the flowers, I want to make sure you can’t argue with me over which kind of tiger it is if I happen to guess wrong.
I happen to be studying lojban at the moment, and I think the designers have defined linguistic ambiguity not as the opposite of specificity (one of the first lojban words I learned was “zo’e” /ZO.he/, which means something like “contains contextually sensitive information that makes this utterance true, the exact value of which is irrelevant or obvious”), but rather as a linguistic property whereby a semantic construct cannot be pinned down to communicating a specific value. The classic English language example is “time flies like a banana”, in which any of the first three words can be the verb.
Whoa, it never occurred to me that
timecould be the verb there.I feel like this is almost a degenerate-case explanation, but perhaps the primary application of ambiguity is brevity.
The Persian king threatens, “If we win this war, we will kill all your men and enslave your women.”
The Laconian king answers, “If.”
The Persian king says, “I’m not sure what you mean by that.”
The Laconian king says: “Like, ‘If you win.’”
The Persian king says: “‘If you win’ what? Like I said, if we win, we’ll kill you and enslave your women.”
The Laconian king says: “Right. If you win.”
The Persian king says: “I’m so confused.”
The Laconian king says: “I’m emphasizing the use of ‘if’ to draw attention to the idea that I don’t plan to let you win.”
The Persian king says: “Ohhhh oh I get it. Okay, yeah, that’s pretty good. Scribe, blot out everything after he said ‘If.’”
As much as I like perfect clarity, brevity does have its uses. When I was reading Atlas Shrugged, I cringed every time a character replied to a comment with “I know it” instead of the normal “I know.”
Note that ithkuil(the constructed language we were talking about in that thread) aims for both clarity and brevity.
Ambiguity isn’t a unary function. It doesn’t make sense to say “this sentence is ambiguous” without some context:
What’s the social context in which the sentence is spoken? “I love you” from a child to a parent almost certainly doesn’t mean the same thing as “I love you” from an adult to a lover. This isn’t ambiguity; it’s context dependence.
What are the likely (mis)interpretations? How distant are they from each other in meaning-space? Do those differences matter? “Go to the library and get me Moby-Dick” might leave the hearer unclear as to which library is meant (the college library? the public library? the main branch of the public library? the room in the house with lots of bookcases?) … but maybe the speaker doesn’t care about the difference — they just want their damn whale book.
Is the sentence intended to nail down one specific meaning when some hearers would prefer a different one? For some purposes, such as laws and rules, people want a very clear idea of what’s being forbidden, even in a context where people have strong disagreements about what should be forbidden. See for instance the U.S. jurisprudence notions of the vagueness doctrine and the “chilling effect”.
For that matter, what level of language skill does the speaker/writer expect from the hearer/reader, and how much care is the hearer/reader expected to apply? Some grammatical forms are harder to parse. There are a number of example sentences where introducing a comma or pause in the wrong place can completely change the sense of the sentence.
“Woman: Without her, man is nothing” / “Woman, without her man, is nothing”.
”I helped my Uncle Jack off a horse” / “I helped my uncle jack off a horse”.
”Unless a player has Book Burning deal six damage to him or her …” / “Unless a player has Book Burning, deal six damage to him or her …”
In the original context, the alleged desirable ambiguity was the ability to concisely omit information—that is, to say “people” instead of “men and women”. Tabooing ‘ambiguity’, I’d frame this as a matter of having words for large sets rather than requiring speakers to construct them out of smaller sets, and say that this is a good thing if those sets are commonly referred to.
On a similar note, there can be intensions whose extensions are not agreed upon—”good” and “right” spring to mind. At first I thought it would be necessary to have words for these, but upon reflection I’m not sure. Could we replace them with more specific words like “right according to classical utilitarianism” or “right according the ethics of the person this word relates to”?
Sounds like the difficulty here is in the difference between ambiguous and vague or general.
lojban handles that by defining words on a fill-in-the-blank basis, where you can leave out any (or all) of the blanks to be less specific. “good”, for example, is defined “*[object/event] is good for [beneficiary] by [standard]*”. A speaker/writer may fill in as many of the details as (s)he feels are necessary to communicate. As the author, you may be as vague about the specifics as you like, but what you actually said always selects for an exact, unique semantic region.
What if you want to say that something is good but that there is no beneficiary and no standard? A language should not exclude certain possibilities just because the person that invented the language thinks that they cannot happen in reality.
By what standard should it not?
If you don’t want to specify a beneficiary or standard, you can just leave those blanks empty, and people often do so, allowing context to fill in the necessary information. If a listener needs more detail they can always ask for it. lojban questions happen to take the form of requests to fill in a specific blank.
I gather that back in the 1990s there was quite the argument regarding what place structure certain words “should” have. In the end, the Logical Language Group decided on the current definitions and declared that they were making no metaphysical claims, it’s just that words have to be defined one way or another. Too, there are grammatical ways to add, explicitly delete, or rearrange places in any word if the author should feel the need.
The result is that the place system is extremely horrible and hard to use because every word has it’s own rules about it’s places.
I’ve not found it to be so, but you are welcome to your own evaluation, friend. :)
In general, x1 tends to be analogous to an English subject and x2 like an English object for gismu where that applies. At any rate, I don’t think of memorizing even 1300 completely random place structures as any great mental burden compared to, for example, conjugating verbs in English, Spanish or French; or declining nouns in Latin.
Yes, if you only use the x1 and x2 of a gismu that’s true. You are also not finished by memorizing 1300 place structures because Lujvo also have places and you can’t be sure which places of the words on which the Lujvo is build get actually used for that purpose.
Fair enough, but still far less troublesome than the structure of any natural language I’ve ever studied, IMO, where each of many thousands of words must be memorized separately for mastery. At least there are consistent rules, even if they are currently incomplete or inadequate in some places. The language is still very young, after all. Even the documentation admits that lujvo are a bit problematic. Room for growth is not necessarily a bad thing.
This is a start toward what may be a better way of communicating (than many natural languages by standard of understandability ;). If nothing else, Lojban can show us some places where our own languages are… less than rigorous in promoting good mental habits. Our languages, cultures, and customs leave us with plenty of blind spots and I am happy for any tool that can shed light on even one.
Room for growth suggests that Lojban could outgrow it’s horrible place system. I don’t think that’s true. It’s a core design flaw. Admitting that lujvo are problematic doesn’t help. They are a core feature of the language.
There’s nothing consistent about gismu/lujvo places. The lojban dictionary isn’t even consistent on the meaning of
melbi/beautiful.The translation to french suggests that X3 is a standard. The translation to English suggests that X3 is aspect while X4 is a standard. From the outside I might think that the pure lojban dictionary is the most important definition and it suggests that the word has three places. On the other hand most of the translations have four places. The English one has four places.
If we take the English one has canonical then why does X3 mean aspect for
beautifulbut X3 for good isn’taspect? That’s far from there being consistent rules.The lack of prepositions means that you can’t use them for backrefrerencing. That leaves Lojban which wants backreferencing to be evaluated automatically with the very ugly way of backreferencing via the beginning letter of words.
You can make the same arguments for a lot of languages. I do grant that Lojban has a variety of unique ideas that are useful to think about when designing a new language but too much is flawed at the core of Lojban for it being more of the toy language that it’s at the moment.
I wrote a few ideas of how a better language can look like in http://lesswrong.com/r/discussion/lw/n0h/linguistic_mechanisms_for_less_wrong_cognition/ . The rise of China and the way China works, suggests to me that they likely will decide in a few decades that they don’t want to do their science in English. It would be great if we have a real alternative by that point that’s acceptable to the Chinese because it’s culturally neutral. I don’t think Lojban can be that language.
You see a fundamentally flawed system. I see resonances with word games played by children and the foundations of computer science. We may be looking at something that touches deeply in our psyche here and that makes it worth continuing to explore.
Have you reported this bookkeeping error to the LLG? I’m sure they would be happy to correct one dictionary or the other if you did.
Lojban offers a robust backreferencing facility. There are the KOhA and BRODA series for starters, and if you don’t like using variables you can always simply name any grammatical construct using [something] du la [name].
Irrelevant as creating an international auxiliary lanugage is not the goal of the LLG. source
I’m happy that you have given the topic so much thought! I sincerely and enthusiastically look forward to studying your final product.
Peace, Friend.
Loglan did try to implement some of computer science but based on a 1955 understanding of what became to be computer science. James Cooke Brown didn’t focus on math. Math had to be created by the Lojban project as an afterthought. Because Loglan isn’t number friendly it gave useful gismu space to months and weekdays instead of simply calling them by numbers. It has 4 cardinal directions like natlangs instead of allowing the user to specify any angle he pleases.
Lojban forget to include graph theory. I layed out how graph theory could be used to describe relationships in another thread
As far as backreferencing goes KOhA give you the ability to say: la .alis. klama le zarci .i ko’a goi la .alis. cu blanu Alice goes-to the store. It-1, also-known-as Alice, is-blue.
The problem with that is that it has to repeat unnecessary information.
I would prefer a construction that says: Alice goes-to the store. “Agent of last sentence” is-blue.
Along with one that says: Alice goes-to the store. “Destination of the last sentence” is-red.
The information about
agentanddestinationis already communicated in a way where the brain has to keep track of it to understand the message.To do that consistently you would need an ontological commitment that the destination for
goes tois the same ontological concept as the destination forruns to. As far as Lojban is concerned those two concepts of destination have nothing to do with each other.5.4) la .alis. goi ko’a klama le zarci .i ko’a cu blanu Alice, also-known-as it-1, goes-to the store. It-1 is-blue.
No, I don’t know how the internal reporting process of the LLG works. The probably also should simply have a script that checks for all words whether the translations have the same
Lojban doesn’t seem to have any real goals. At the present it’s a toy language and it will likely stay that way. On the other hand it would be great to have a useful loglang that takes a bigger role in society.
It will probably take 5 to 10 years. On open problem is language governance. How do you create a way that multiple people can work at the same language together while with minimal commitments to starting concepts?
I lack understanding of how new word formation works in languages that derive everything from roots like Esperanto/Hebrew/Chinese. It seems that new body of knowledge in academic science but also other fields of inquiry need a way to generate new words. CFAR’s coinage of
a tapfora trigger-action planseems to be a good move to bring a new concept into the language.Apart from CFAR language development I also engage with other alternative knowledge systems build on top of English/German/French. I personally am not big on writing poety directly but I explore overloading words in other contexts like hypnosis.
I think open discussion about the flaws of existing languages is very important even when not immediately proposing an alternative.
Oh, you want ri, ra, ru from selma’o KOhA5, and the go’a series from selma’o GOhA. You can also just use ko’a without explicit assignment and trust the audience to get the meaning from context the same way we do in most natural languages.
la .alis. klama le zarci .i ra goi ko’a cu blanu .i ko’a cu sidju
The ra selects for la .alis. without having to repeat any information. Alternately, if you don’t trust your audience to understand counting rules, la .alis. goi ko’a cu klama le zarci .i ko’a cu sidju mi works just fine as well.
The problem is not “understandign counting rules” the problem is that it takes mental bandwith to do counting. It doesn’t take mental bandwith to know which of the objects in the last sentence was the destination.
On the other hand it’s mentally easier to calculate the distance between month-3 and month-6 then between March and June. But that kind of thought didn’t enter into Loglan. It just copied the way Western languages talk about months.
It’s easy to take a dictionary and simply give every English concept a new name. It’s quite hard to actually think at the fundamental level about the concept in question and how reality can be sliced. Jahai seems to slice odors better than English. But it might be not trival to teach Jahai derived odor categories.
Maybe for ordinary people (like me), the mental difference between March and June is not just the number of days separating them, and does not equal MD between, say, July and October. Calling them by numbers would mean having to re-tie the connotation tails to new symbols, nothing more.
Your discussion points are all over the map. We’ve gone way off the topic of ambiguity vs. specificity. It sounds like you’ve got an ax to grind against Lojban, and now Loglan as well. I’m not here to help you with that.
fe’o
I think it’s worthwhile to critize the flaws of existing languages. I don’t have a specific ax to grind against Lojban. It’s an interesting experiment that provides data.
It’s a more interesting language to learn from then from a language like ROILA which is much worth thought out.
Kinda reminds me how in Tatar language endings of nouns can specify direction and other stuff (kitaphanede—in the library.)
I’m afraid that really only holds for klama (X1 goes to X2 [destination] from X3 [origin] via X4 [route] using X5 [transportation]) and, perhaps, a few other specific words due to their place structures. There is an extensive set of particles (cmavo) to discuss location and directionality in (usually up to four) dimensions. These, or compounds made from them, may be appended to many grammatical structures. Though, unlike many natural languages, lojban does not actually require utterances to make claims about time or number.
The reference material for the lojban space/time system is here if you are interested, and a quick-reference sheet here, though the latter lacks the grammatical explanations that help the system make sense.
You might expand already existing words like “profitable” and “fitting” for those respective ends.
Ambiguity is good for plausible deniability, among other things...
Ambiguity has its uses. In flirting for example, where the offer of intimacy is made implicitly so rejection can be given implicitly (by pretending not to have heard the offer) in order to avoid the pain of explicit rejection.
And in all sorts of situations where explicit statements of particular facts or opinions are punishable, the ability to make them in ambiguous, i.e. deniable form, can be crucial.
But I find the baseline of unintentional ambiguity in most forms of human-to-human communication to be a problem, not an asset, most of the time.
No. Yes.
I interpret “ambiguity” as “having multiple meanings in the given context”. It can be a bad thing or a good thing. The “optimal level of ambiguity”, I think, is very much situation-dependent and I’m not optimistic about formulating any general rules for that.
Off the top of my head I can come up with a couple of way ambiguity is useful. One is when you are thinking at several levels simultaneously. In such a case ambiguity can provide vital linkages between levels helping to keep them in sync. Two is when you use ambiguity to postpone choosing a path for a bit. Here it’s similar to a wavefunction which you might find convenient to hold in an uncollapsed state for while and then collapse it (= resolve the ambiguity) at the right moment.
https://en.wikipedia.org/wiki/Cooperative_principle
We have to make some assumptions that communication is being done in good faith. otherwise all communication would suffer.
some of what you are asking about is covered in that principle.
I think some ambiguity is useful for communication; but at some point people should be making specific claims which can then be discussed and proven true/null etc. (i.e. both are needed)
In general when you refer to another thread on LW, it’s good to add a link.
How do I do that? Is it sufficient to edit a web address into the text of the old and new threads?
Yes.
If you want to put the link behind LessWrong in a comment you have to write [LessWrong](http://lesswrong.com). I think for the topline post it’s html LessWrong
Done. Not pretty, but the links seem to work.