For some reason, I keep thinking about what General Semantics might be trying to get at. In my previous post on the subject, I identified equivocation as the main curse which General Semantics seeks to provide a counterspell for. (AKA conflation, identification, fusion, …) However, equivocation has a lot of faces. It doesn’t work to just say “try really hard not to equivocate”—we need the details. We need tools to notice specific classes of equivocation and un-equivocate them.

Today, I want to talk about equivocation which conversational implicature brings about.

Conversational Implicature

In case you’re not familiar with the term, “conversational implicature” refers to all the meaning that’s not explicit. For example, if I say “you didn’t do the laundry yet?” I might (depending on context) be implying “you should really do it now”. This is implicature, because I didn’t say it explicitly; it is inferred from context.

I’ve written before about Bayesianism and the explicit/implied distinction. I wrote that Bayesian signalling theory isn’t adequate to understand the distinction: meaning is meaning, and in particular, the meaning of an utterance is what people understand from it. How could we make a Bayesian distinction between “explicit” and “implied” meaning?

(To put it a different way: from a map/territory standpoint, someone talking about “explicit” meaning might sound confused: the meaning of the map is precisely the way it corresponds to (IE statistically conveys information about) the territory. It’s not like you actually hand over chunks of the territory! So what is this “explicit meaning”?)

I nonetheless do think there is such a thing as “explicit meaning”. It may have gradations and nuances (many “explicit” phrasings have their roots in analogy, eg, “she’s on time” doesn’t make sense if you try to be very literal about what “on” means), but by and large, we can make a reasonable dividing line.

Policy Debates

In Policy Debates Should Not Appear One-Sided, Eliezer identifies a common failure mode, where listing cons is conflated with arguing against all the pros, (and listing pros is conflated with arguing against all the cons):

Robin Hanson proposed stores where banned products could be sold.¹ There are a number of excellent arguments for such a policy—an inherent right of individual liberty, the career incentive of bureaucrats to prohibit everything, legislators being just as biased as individuals. But even so (I replied), some poor, honest, not overwhelmingly educated mother of five children is going to go into these stores and buy a “Dr. Snakeoil’s Sulfuric Acid Drink” for her arthritis and die, leaving her orphans to weep on national television.
I was just making a factual observation. Why did some people think it was an argument in favor of regulation?

I now think this is one example of the more general rule: people conflate explicit content with conversational implicature all the time.

In a policy debate, it’s common that there is one big decisive factor. So it often makes sense for participants to bring up things they think are decisive in one direction or the other. So when someone brings up a pro/con, there is, statistically, an implicature that the pro/con is decisive.

Of course, this is expected, and usually beneficial. Implicature is part of the meaning of an utterance. It’s socially expected that you’ll understand it and respond to it.

In fact, I think it’s often useful to totally ignore explicit content and deal only with implicature (this reduces friction when dealing with people who are not very rationalist-compatible).

But carrying this too far is toxic for policy debate. If every pro/con is seen as implicitly carrying the claim “this is decisive”, then anyone listing a pro/con is seen as arguing against all the cons/pros, which makes cost/benefit analysis impossible.

Against Implicature?

I used to derive a heavily anti-implicature lesson from Eliezer’s post. I said to myself: it’s important to sometimes make true observations without a “point” you’re arguing for/against; otherwise you will have already written the bottom line, so you won’t be doing useful computation.

But this is a bit absurd. Does the ideal rationalist just spout facts at random, to avoid drawing a bottom line?

Similarly, one might read Eliezer as suggesting that people shouldn’t make implicature inferences. “I never mean that a con is decisive, unless I say so, and you should never infer that about someone else either!” But this is almost as absurd. Ignoring implicature is not going to be practical.

Just Stop Equivocating

Instead, the advice I derive is try to stop conflating the explicit content and the implicature. This doesn’t mean valuing one and discarding the other. Don’t think like one is “more legitimate” than the other. Just note the difference between them.

There’s a weird meta thing going on here: if I say “note the difference”, I worry you’ll hear some specific point to it, like “note the difference so that you can discard the implicature” or “note the difference so that you can pay attention to the implicature”. No! There is no one appropriate rule for all circumstances.

STUDENT: Then why do we note the difference? Surely there is some point?

MASTER: Yes, of course there is. I want you to stop equivocating because equivocation is dangerous. We already covered how toxic it can be for policy debate.

STUDENT: I understand that example, but what general lesson am I supposed to learn? Surely you want me to stop equivocating because you expect good consequences. But what are the gears of those good consequences? What do you expect to happen next, if in a particular case I succeed at “stop equivocating”?

MASTER: In some cases, you will realize that the literal content was itself the point, and you should stop searching for one. Imagine a Master of Literalness who always speaks literally to the student, and tries to teach the student to do the same. Every day, the master lectures on speaking literally. Every day, the student asks the same question: what is the point of all this, master? When asked the point, the master only repeats the same teaching, over and over again. The student despairs. One day, the student realizes that the master was speaking simply all along. The point was precisely what the master was saying. The student is enlightened.

STUDENT: But this is not your teaching. You don’t tell us that we should speak simply at all times.

MASTER: No, it is not my teaching. The student in the story learned a valuable lesson, namely, that it is possible to speak completely literally with no hidden agenda. However, possible does not mean desirable. The point of the story is that the student was unable to see what was right in front of them, because they thought every sentence should have a hidden meaning, a “point”. This can happen to you if you live in the world of implicature alone. Other bad things can happen to you if you live in the world of literal meaning alone.

STUDENT: So you’re saying that we should consider the possibility that people mean exactly what they say.

MASTER: Yes, but also consider the possibility that they don’t.

STUDENT: So we should consider all possibilities at all times.

MASTER: Humans are not capable of this. We need to ignore most of the possibilities most of the time. But this creates a trap: how can we learn that we need to raise something to our attention, if we’re always ignoring it? I’m mentioning the possibility now, so that you might raise it to attention when it’s relevant, and avoid the trap.

But how do we stop equivocating?

Actually, “just stop equivocating” isn’t very specific advice. If you’re equivocating, it can be really, really hard to stop.

I think it’s a bit like grammar. You could speak English fluently all your life, but not be able to identify a “preposition” or a “gerund”. Just because your brain is juggling explicit meaning and implicature all the time doesn’t make the distinction readily available.

But it’s even worse than grammar, because with grammar, everyone goes through multiple semesters of explicit content. I think most people are mostly blind to the literal/implicature distinction most of the time (especially young people?). People know it exists, but only really think of it as applying to extreme cases (such as veiled threats). More knowledgeable people might know that implicature happens “everywhere all of the time”, but think you have to be a linguist to spot it.

So, if you had a blind spot around the distinction, but wanted to get better, how would you start?

It’s helpful to just pay attention. What did you say, and what did you mean by it? What did other people say, and what do you think they meant by it?

It’s helpful to have peers who are also doing this, and talking about it.

I think it’s easiest to notice when someone infers something which you didn’t intend. Then, the subject can be brought up and examined. (Don’t attack the person with “that’s not what I said”! Remember, inferring implicatures is a perfectly normal part of functional communication. Scolding people for it only serves to reinforce implicature blindness. The gentler “that’s not what I meant” makes more sense.)

The English Language

The English language explicitly conflates the two in the phrase “what are you saying?”—I often get confused reactions when I comment about “what a person said” (intending to refer only to the explicit content). Even when I make the distinction clear, I often get reactions like I’m playing some verbal game or something. (I think people might usually make explicit/implicit distinctions as a joke!)

The evolution of the term “literally” into an intensifier doesn’t help matters, obviously. It’s already a bit sad when a word like “impossibly” has its literal meaning eroded as it gains use as an intensifier (as in the phrase “impossibly large”, when used to refer to things which are quite possible). It’s darkly ironic when it happens to the word we would use to distinguish between such exaggerations and plain language!

Speaking Plainly

Rationalists are, on the whole, probably considerably above average in literalness. I’ve argued that one should not try to go all the way: implicature is good and proper.

But could one go all the way? As an exercise, perhaps?

I would argue that, yes, it is possible to say almost all of what you mean. It’s possible to pause a conversation and unpack, and within a few minutes, to be finished unpacking to the point where any further details are negligible. Perhaps you can’t recursively unpack everything (EG, getting into your own intentions behind pausing to unpack); that might get into an infinite recursion. But I claim you can unpack any one thing with reasonable thoroughness in a reasonable amount of time.

(My own explorations of this have been under the influence of a drug, which made it feel necessary to unpack implicature to literal content in order to be sure of what’s going on; so, take that as you will.)

In many cases, this is accomplished by replacing implicature with vague language; EG, “The Latin language doesn’t rely so much on word order and that’s somehow good” (without trying to immediately unpack what ‘good’ means). That’s OK. It can open the door for further clarifications if needed. We don’t have to resolve all linguistic ambiguity at once.

This takes some of the same skills as Radical Honesty. Often, there’s a good reason (beyond efficiency) why so much is left unsaid. Saying them directly is perceived as too forceful, impolite, etc. It can take courage to speak literally.

Imagine for a moment that you are a student of the Master of Literalness from my earlier story. You’ve been soaked in pro-literalness culture. To you, leaving things unstated feels dishonest and dangerously ambiguous. You know the importance of one’s words, and you see how people constantly avoid culpability by saying things without saying them. You say what you mean.

It seems like an interesting exercise, to try and talk as if you had such an extreme perspective.

There’s a lightness that comes with speaking plainly. You don’t have to defend some hidden, unarticulated point. If someone says something true, you can simply agree with them.

But in the end, we have to wake back up to the world of implicature. If someone expresses (what they think is) a counterpoint to what you are saying, and you simply agree, they’ll probably think you concede the whole argument. Maybe you should! But if not, you should probably recognize the implicature and attempt to communicate further.

Implicature Conflation