Miscellaneous Musings on Language

What does it mean to say “She motioned me to come forward.” This is the thing, above every example I have seen, that should make us suspicious about any definite notion of “meaning”.

The idea of meaning started as arguments over a ledger—the ledger is what people agree has been spoken. Previously, to win arguments people were fond of arguing over the symbols on the ledger—what was said—and of course this still happens even with airtight recordings. But in general the conversation has shifted to what things mean and this argument has created a massive demand for technologies of equivocation. Of course, they’re not called equivocation, because they’re coordinated around group identity, so within the group things are not only clear, they’re more precise than they used to be because they are honed to pass specific messages to certain people and take advantage of intersubjective agreement. But in truth, this creation is one of getting around the definiteness of the ledger, to allow people room to breathe, for human beings were never meant to be bound to an irreversible ledger.

What should we try to describe, when we talk about language? A naive view would be say we should be able to predict what people say. This is clearly impossible, simply due to information access that no one will have in the near future. Of course, our confidence about what someone will say should certainly go up, but probably most of that is being done pretty well by GPT-3 and the correct calibration of the word “the” will probably be more important than anything we write down for the purpose of increasing density….

I would argue that our goal is to get the mechanics down in a way we can do something with, and the question should be what do we want to do?

Well, one thing I’d like to do is argue about memetic strategy—why do people make explicit cultural references in order to garner certain kinds of attention, support, and backlash? How do they know what they are going to get? Of course people don’t know but there are plenty of good examples of people who are incredibly effective at memeing their part of the internet into a certain shape. How can we formalize their knowledge?

This is one of the easiest things to do, because of data access, but when it comes down to fundamental questions these kinds of analyses become even more important in face-to-face relationships. People shrug, purposefully misunderstand, react in degrees, choose to follow certain paths, talk quickly, move around, bring-up parts of the physical environment, use body language to adjust their implied attention, and a whole host of other moves in order to structure conversation. What are they accomplishing? It might be tempting to try a bottom-up description of this: show how facing someone shows attention and then describe what that means, but the top-down compositionality is too strong for me to take this seriously. In fact, communication has so much top-down compositionality I’m convinced one of the major draw backs is that we can’t think of reasonable examples to share in our powerpoints that people will think are polite, be able to understand, and be willing to agree with in public.

We might not know exactly what to explain, but whatever it is, it should explain why a definitive choice is taken. You’ll know a definitive choice when you observe it, by the fact that it deviates from a counterfactual version that wouldn’t have drawn your attention. We don’t and can’t access these counterfactuals and even finding many copies that are similar enough that we can use lots of data to control for the differences is implausible. That’s why this is the Inexact Sciences—we need to rely on intuitions, assume they are true and try to explain the cracks inbetween. But we will not specifically “accept” these intuitions, only confirm them with an extrinsic evaluation: our ability to explain other emergent behavior that would result, e.g. how the nascent ceremony of one generation will interact with another.

Where should we begin? We should begin by thinking on what intuitions we do trust or else face the need fortabula rasa science—a prospect I think is not just bad, but truly impossible. Hypotheses come from somewhere. In a space where we cannot execute reproducible experiments, we must especially rely on intuitions, because we can only test thoroughly at higher levels of abstraction (aggregate data of social experiments) and lower levels of abstraction (lab experiments that deal with cognitive effects).

Some intuition:

  • Ceremonies generally manipulate underlying objects that serve social roles.
  • Complaints are requests for sympathy and sometimes call for systematic change—they serve to establish what is a truth and what is personal by the reaction of the community.
  • Their decay indicates these objects decaying or a new game that manipulates these objects, a key signal to causality.
  • Compliments make people happy, but their strength in a given context indicates their sincerity.
  • People seek status.
  • Status is naturally produced by scarcity hubs embodied in well-connected and highly-appraised people.
  • In the modern age, defining boundaries of in-groups at multiple different levels is a key purpose of communication.

Let us finally begin in earnest.

Published by Crispy Chicken

Take a bite.

%d bloggers like this: