= 27d567bc2d48d9f40e9ccc20ea370b15)
RichardKennaway comments on Diseased thinking: dissolving questions about disease - Less Wrong
You are viewing a comment permalink. View the original post to see all comments and the full post content.
= 783df68a0f980790206b9ea87794c5b6)
Loading…= b715d02e85f7b3b4ae65ca3d3e450868)
Subscribe to RSS Feed
Powered by Reddit
= f037147d6e6c911a85753b9abdedda8d)
You are viewing a comment permalink. View the original post to see all comments and the full post content.
Comments (343)
In the least convenient possible world, condemning an innocent in this one case will not make the system generally less worthy of confidence. Maybe you know it will never happen again.
Maybe everyone would have a pony.
ETA: It is not for the proponent of an argument to fabricate a Least Convenient Possible World -- that is, a Most Convenient Possible World for themselves -- and insist that their interlocutors address it, brushing aside every argument they make by inventing more and more Conveniences. The more you add to the scenario, the smaller the sliver of potential reality you are talking about. The endpoint of this is the world in which the desired conclusion has been made true by definition, at which point the claim no longer refers to anything at all.
The discipline of the Least Convenient Possible World is a discipline for oneself, not a weapon to point at others.
If I, this hypothetical judge, am willing to have the innocent punished and the guilty set free, to preserve confidence that the guilty are punished and the innocent are set free, I must be willing that I and my fellow judges do the same in every such case. Call this the Categorical Imperative, call it TDT, that is where it leads, at the speed of thought, not the speed of time: to take one step is to have travelled the whole way. I would have decided to blow with the mob and call it justice. It cannot be done.
If that's what makes the world least convenient, sure. You're trying for a reductio ad absurdum, but the LCPW is allowed to be pretty absurd. It exists only to push philosophies to their extremes and to prevent evasions.
Your tone is getting unpleasant.
EDIT: yes, this was before the ETA.
I think you replied before my ETA. The LCPW is, in fact, not allowed to be pretty absurd. When pushed on one's interlocutors, it does not prevent evasions, it is an evasion.
The categorical imperative ignores the possibility of mixed strategies--it may be that doing X all the time is bad, doing Y all the time is bad, but doing a mixture of X and Y is not. For instance, if everyone only had sex with someone of the same sex, that would destroy society by lack of children. (And if everyone only had sex with someone of the opposite sex, gays would be unsatisfied, of course.) The appropriate thing to do, is to allow everyone to have sex with the type of partner that fits their preferences. Or to put it another way, "doing the same thing" and "in the same kind of case" depend on exactly what you count as the same--is the "same" thing "having only gay sex" or "having either type of sex depending on one's preference"?
In the punishment case, it may be that we're better off with a mixed strategy of sometimes killing innocent people and sometimes not; if you always kill innocent people, the justice system is worthless, but if you never kill innocent people, people have no confidence in the justice system and it also ends up being worthless. The optimal thing to do may be to kill innocent people a certain percentage of the time, or only in high profile public cases, or whatever. Asking "would you be willing to kill innocent people all the time" would be as inappropriate as asking "would you be willing to be in a society where people (when having sex) have gay sex all the time". You might be willing to do the "same thing" all the time where the "same thing" means "follow the public's preference, which sometimes leads to killing the innocent" (not "always kill the innocent ") just like in the gay sex example it means "follow someone's sexual preference, which sometimes leads to gay sex" (not "always have gay sex").
Yes, the categorical imperative has the problem of deciding on the reference class, as do TDT, the outside view, and every attempt to decide what precedent will be set by some action, or what precedent the past has set for some decision. Eliezer coined the phrase "reference class tennis" to refer to the broken sort of argumentation that consists of choosing competing reference classes in order to reach desired conclusions.
So how do you decide on the right reference class, rather than the one that lets you conclude what you already wanted to for other reasons? TDT, being more formalised (or intended to be, if MIRI and others ever work out exactly what it is) suggests a computational answer to this question. The class that your decision sets a precedent for is the class that shares the attributes that you actually used in making your decision -- the class that you would, in fact, make the same decision for.
This is not a solution to the reference class problem, or even an outline of a solution; it is only a pointer in a direction where a solution might be found. And even if TDT is formalised and gives a mathematical solution to the reference class problem, we may be in the same situation as we are with Bayesian reasoning: we can, and statisticians do, actually apply Bayes theorem in cases where the actual numbers are available to us, but "deep" Bayesianism can only be practiced by heuristic approximation.
"Would you like it if everyone did X" is just a bad idea, because there are some things whose prevalences I would prefer to be neither 0% nor 100%, but somewhere inbetween. That's really an objection to the categorical imperative, period. I can always say that I'm not really objecting to the categorical imperative in such a situation by rephrasing it in terms of a reference class "would you like it if everyone performed some algorithm that produced X some of the time", but that gets far away from what most people mean when they use the categorical imperative, even if technically it still fits.
An average person not from this site would not even comprehend "would you like it if everyone performed some algorithm with varying results" as a case of the golden rule, categorical imperative, or whatever, and certainly wouldn't think of it as an example of everyone doing the "same thing". In most people's minds, doing the same thing means to perform a simple action, not an algorithm.
In that case, the appropriate X is to perform the action with whatever probability you would wish to be the case. It still fits the CI.
Or more briefly, it still fits. But you have to actually make the die roll. What "an average person not from this site" would or would not comprehend by a thing is not relevant to discussions of the thing itself.
In that case, you can fit anything whatsoever into the categorical imperative by defining an appropriate reference class and action. For instance, I could justify robbery with "How would I like it, if everyone were to execute 'if (person is Jiro) then rob else do nothing'". The categorical imperative ceases to have meaning unless some actions and some reference classes are unacceptable.
That's too brief. Because :"what do most people mean when they say this" actually matters. They clearly don't mean for it to include "if (person is Jiro) then rob else do nothing" as a single action that can be universalized by the rule.
The reason that doesn't work is that people who are not Jiro would not like it if everyone were to execute 'if (person is Jiro) then rob else do nothing', so they couldn't justify you robbing that way. The fact that the rule contains a gerrymandered reference class isn't by itself a problem.
Does the categorical imperative require everyone to agree on what they would like or dislike? That seems brittle.
This post discusses the possibility of people “not in moral communion” with us, with the example of a future society of wireheads.
I've always heard it, the Golden Rule, and other variations to be some form of "would you like it if everyone were to do that?" I've never heard of it as "would everyone like it if everyone were to do that?". I don't know where army1987 is getting the second version from.
Doing which is reference class tennis, as I said. The solution is to not do that, to not write the bottom line of your argument and then invent whatever dishonest string of reasoning will end there.
No kidding. And indeed some are not, as you clearly understand, from your ability to make up an example of one. So what's the problem?
What principle determines what actions are unacceptable apart from "they lead to a bottom line I don't like"? That's the problem. Without any prescription for that, the CI fails to constrain your actions, and you're reduced to simply doing whatever you want anyway.
It's not like the issue has never been noticed or addressed:
"Hypothetical imperatives apply to someone dependent on them having certain ends to the meaning:
if I wish to quench my thirst, I must drink something; if I wish to acquire knowledge, I must learn.
A categorical imperative, on the other hand, denotes an absolute, unconditional requirement that asserts its authority in all circumstances, both required and justified as an end in itself. It is best known in its first formulation:
Act only according to that maxim whereby you can, at the same time, will that it should become a universal law.[1] "--WP
This asserts a meta-meta-ethical proposition that you must have explicit principles to prescribe all your actions, without which you are lost in a moral void. Yet observably there are good and decent people in the world who do not reflect on such things much, or at all.
If to begin to think about ethics immediately casts you into a moral void where for lack of yet worked out principles you can no longer discern good from evil, you're doing it wrong.