Response to: Conversation Deliberately Skirts the Border of Incomprehensibility (SlateStarCodex)

“Why don’t people just say what they mean?”

They can’t.

I don’t mean they don’t want to. I don’t mean that they choose not to. I don’t even mean they have been socially conditioned not to. I mean they literally have no way to do the thing. 

I don’t even mean that they don’t know how in some sense, that they have not realized the possibility. It is literally impossible to “just” say what you mean. With work, you can mostly just say what you mean, but even that is hard. 

This is because humans are not automatically strategic or automatically scientific but by default humans are automatically doing lightning fast implicit probabilistic analysis on social information in the background of every moment of their lives. Most of them would not describe it that way, but that is a thing that is going on.

So let’s take the example that starts the quote in Scott’s article: Bob is watching a television show and Alice asks what he is watching.

My experience says that most of the time, this primarily really is a request for information. Alice wants to know what show Bob is watching.

It also raises the probability that Alice will want to watch the show with Bob, conditional on what show Bob is watching; there is even a chance that Alice knows what show it is, and is using the question to open the conversation for this purpose. Certainly, this would be a reasonable thing to do, were Alice potentially interested in watching with Bob. Alice could also simply be making conversation, and want something else entirely. For all we know, she just wants to make a new friend, or plans to distract him and stab him with a knife.

Alice is providing at least very weak evidence in favor of those possibilities, and many others, while providing evidence against everything else.

That doesn’t mean Alice wanted to send all of those messages, but whichever one Alice had in mind, there was no way for her to send only that message – any other thing she could have said would also have had lots of likelihood ratios attached to it.

Then, in turn, Bob’s response does all of these same things. Telling Alice what the show is gives her that information, which may or may not have been what she wanted. It also indicated that Bob was willing to talk to Alice and to spend minimal effort to give her information, and the tone and detail of how he did that will also be information. It also means Bob turned down the opportunity to interact in some other way, which is information.

The majority of the time, Alice primarily wanted to know what Bob was watching, but to find out, she had no choice but to send out a message with all these other implications as well. Bob was happy to provide the name of the show, but in doing so is also leaking tons of information, no matter how he does it.

There is no such thing as someone just saying what they mean in a normal social context. None.

None of that requires any ulterior motives on Alice or Bob’s part, either selfish or selfless. Alice and Bob can both want to have the literal ask-then-answer interaction on its own level with no implications, but that only happens if at least one of them is blindly throwing out tons of evidence that humans do not naturally throw out.

If Alice is sufficiently blind to social cues and these sorts of implications (no need to put a label on this) and Bob knows this then Alice can ask not knowing about the other implications of the question, and Bob can know that Alice doesn’t know those implications, which in turn means that Bob only has to update on the information that Alice wants to know what show Bob is watching enough to ask Bob for that information, and Bob can then answer knowing that Alice won’t update on the fact that Bob is answering, so Bob doesn’t have to worry about the other implications of what Alice is saying, so no information leaks from either side and you can have a ‘clean’ transaction born of a combination of ignorance and knowledge of that ignorance (the reverse case, or both being blind to cues, also work).

Mostly. You still can’t actually get the name of the show without Bob updating on the fact that Alice asked the question, and Alice updating on Bob being willing to answer. You can only contain that effect.

That’s the best case scenario for isolated communication, where everyone involved wants it to happen, and isn’t worried at all about putting someone else in an awkward situation or needing to make sure the other party has an easy way out or what not. In the normal case you are completely screwed.

Of course, when I say completely screwed, I mean forced to engage in all of this hyper-rich communication with lots of detail and implication that allows the parties involved to accomplish a lot with few words and little time. If all of the moves available to you chance a thousand variables, that is both blessing and curse, and the better you know the game and the other players, the more the game becomes efficient and positive-sum.

The option you do not have, unless you want to take your ball and go home, is not to play.