Tags
The misuse of certain words in the rationalist community drives me up the fucking wall, and so in this post I shall explain the correct usages.
Precommitment. Precommitment is the use of a commitment device to ensure that you will face severe negative consequences from making certain choices you wish to commit to not making. Beeminder is a precommitment device, because if you don’t reach your goals, you will have to pay money.
Does Not Mean The Same Thing As: “Commitment”. If you stood up in front of your friends and family and solemnly promised not to divorce, you’re committed; if you’re going to pay two-fifths of your income in alimony, you’re precommitted.
Cooperate/Defect: These terms are used in a wide variety of different game theory situations, but the one most popular among people who aren’t game theoreticians is the Prisoner’s Dilemma. In the classic Prisoner’s Dilemma, two prisoners are separated. If one snitches on the other, he will get a reward and the other will be imprisoned; if both stay silent, they will both go free; if both snitch, they are both imprisoned. Many situations can be modeled as Prisoner’s Dilemmas.
Does Not Mean The Same Thing As: “Prosocial”/”antisocial”. If you wear an “I cooperate on prisoner’s dilemmas” shirt, I’m going to get worried that you’re planning on starting a wage-fixing cartel.
Schelling Point: The point on which people will naturally converge in the absence of communication. For instance, if you have to meet a stranger in NYC without communicating with them, you will probably meet them at noon at the information booth at Grand Central Station.
Does Not Mean The Same Thing As: “The place at which we agreed to meet each other through explicit communication.” If you communicate with each other, it is not a Schelling point.
Also Does Not Mean The Same Thing As: “Schelling fence“, which refers to the place where people draw the line on a slippery slope. For instance, in the US, the Schelling fence about government intervention in free speech is “no fighting words, obscenity, libel/slander, true threats, or child porn”.
Tom Ash said:
If someone wears an “I cooperate on prisoner’s dilemmas” shirt, it makes me think they cooperate with others only because they see it as in their self-interest – i.e. that they’re not especially pro-social.
LikeLiked by 2 people
The Smoke said:
How do you arrive at that opinion? Unless it’s an iterated prisoners dilemma it is a-priori pro-social to cooperate. (there are no consequences for third parties in a real prisoners dilemma)
LikeLike
davidmikesimon said:
Maybe they just want *you* to cooperate because you think they still. Then the knife goes in…
LikeLike
Blake Riley said:
Standing up in front of your friends and family and solemnly promised not to divorce could be precommitment. It depends on how likely friends and family will impose social punishment for divorce.
Along these lines, it irks me how ‘signaling’ is misused. Signaling theory depends on actions having different costs for different groups so one can’t or doesn’t find it worthwhile to mimic the other. Otherwise, it’s cheap talk or plain old communication.
LikeLiked by 1 person
Max H said:
What you’ve described as precommitment (“Ensuring that you will face severe negative consequences from making certain choices”) is actually just a method for achieving precommitment. For example, an army burning a bridge or removing your steering wheel in a game of chicken are other methods of precommitment that make certain choices impossible, rather than just change their utility.
I think the essence of precommitment is that it allows you to make certain decisions before they become difficult, like deciding in general not to give into blackmail before you’re actually being blackmailed. You can use precommitment to enforce decisions like that, but it doesn’t necessarily require imposing external negative consequences.
LikeLiked by 4 people
Muga Sofer said:
Another possibility is having the sort of mind design that can bindingly commit to future actions.
This is probably where the rationalist usage of the word comes from; discussion of how a Perfectly Rational AI would want to be the type-of-mind that can just say “I commit not to do that thing” and be believed because we can see in it’s code that it literally can’t betray us.
(Well, that plus the Timeless Decision Theory thing.)
LikeLike
geekethics said:
I note a lot of people seem to slip from “an ideal AI would have such a mind” and “I understand TDT” to “I have such a mind”.
LikeLike
davidmikesimon said:
Does deciding in advance to one-box on Newcomb not count as pre-commitment then? I thought that was the textbook example.
LikeLike
Tarn Somervell said:
Your definition of pre commitment is overly narrow though – any way of genuinely commuting to a future choice or action qualifies. Granted, in many cases just saying you will doesn’t do that, but for some people it may.
LikeLike
lahwran said:
I like that we’re talking about this. I consider prescriptivism to be useful when you need to communicate with a group that already uses words a different way; and in that same spirit, can we talk about “rationality”? I haven’t figured out how to say it eloquently yet, but there’s something very *weird* about saying “the art of rationality” and meaning “the thing cfar is making”, or talking about how one “uses rationality”.
LikeLike
sullyj3 said:
Ooft. Guilty.
LikeLike
Murphy said:
“wage-fixing cartel.”
…. you mean a union?
LikeLike