Anyway, if from seeing the overwhelming success of a certain decision, I infer that I would like to also make that decision … No that doesn’t seem ridiculous to me.
People taking IQ tests = Two-boxers
Psychological troubles = being poor (winning less)
Do you genuinely not see that it is the same reasoning?
So you agree with the following sentence: “you should avoid taking IQ tests to not get psychological troubles”?
Maybe a little bit, but it feels strained and irrelevant.
Its not irrelevant. It feels strained because its wrong and you cant accept it
how is it strained or irrelevant? replace the words and it’s the same thing
I don’t think so. There’s no being there modelling your behaviour and choosing what happens based on how it predicts your behaviour. I think that’s a pretty substantial difference.
That doesn’t matter, that’s not the argument you use. Your argument doesn’t need the predictor to do anything crazy. Just that it has a good track record. I quote:
The predictor is correct overwhelmingly about if people one box or not, that’s a given in the problem statement. That necessarily means two boxers overwhelmingly leave with 1000 and one boxers overwhelmingly leave with 1 million.
You can spin “correct overwhelmingly” if you want, the fact of the matter is that you can have the same thing with correlation.
Which is only possible if some agents are not perfectly rational agents, because all perfectly rational agents will make the same choice; if it is more rational to choose one box then all perfectly rational agents will choose one box and if it is more rational to choose two boxes then all perfectly rational agents will choose two boxes.
Not at all. You’re ignoring stochastics for a start. It might, for example, be rational to do the opposite of whatever the last person did. Or it might be rational to mimic randomness (assuming actual randomness to be impossible). It might be rational to pick differently depending on the day of the week, or the declination of Venus…
All we have is that out of 100 trials, the being gets one prediction wrong. We don’t know why.
As I said above, our sensible priors ought be that this is a trick (rather than that causality is broken, or time-travel possible), and that we ‘win’ when we’re in the minority (otherwise it’s a rubbish trick).
Therefore our only rational course of action is to make our best guess as to the cause of the 1% failure and exploit it.
Remember that there are four possible outcomes… You are in the 1% if you are in scenario 4, but you don’t want to be in scenario 4. So it’s not just enough to trick the predictor; you specifically want to trick the predictor into scenario 1.
As I said above, our rational priors ought to be that this is a trick, and that the majority loose the trick. That’s how tricks work.
It’s simply not possible to employ probability reasoning in an unknown probability space using frequentist probability. You’ll end up with whatever result you’ve arbitrarily filled out your probability space with.
The only sensible way to approach probability judgements with unknown parameters is with Bayes. And that involves realistic priors.
I don’t know what “have the same thing with correlation” means
It means the prediction doesn’t cause you to pick one or two boxes in any way so it’s all correlation.
I don’t know what you’re arguing about with that. Who said the prediction causes your choice? Or that your choice causes the prediction? I’m not saying that.
I’ve confused matters by invoking free will as it’s such a deep topic. I just meant to invoke a decision point. Ignore that we can do so freely if it helps. The point is simply that we can do so after the caveat that we all have the same knowledge.
Ok. So the choice doesn’t cause the prediction, right? it’s just highly correlated.
Just like the IQ tests don’t cause the psychological troubles, it’s just highly correlated.
So your argument also applies to the IQ test situation.
I don’t really feel any draw to talk about the iq thing lol.
As an aside, there’s an interesting paper on a Decision Theory known as logical decision theory, or functional decision theory, and one of its selling points is that it produces higher-value choices on a lot of decision theory problems, one of them being Newcombs Problem. It discusses causal decision theory as well when looking at Newcombs Problem. I’ll link to it.
I don’t really feel any draw to talk about the iq thing lol.
I am really trying to comprehend, do you genuinely not see that you are saying that correlation is causality? even if you ignore the IQ example.
As an aside, there’s an interesting paper on a Decision Theory known as logical decision theory, or functional decision theory, and one of its selling points is that it produces higher-value choices on a lot of decision theory problems, one of them being Newcombs Problem. It discusses causal decision theory as well when looking at Newcombs Problem. I’ll link to it.
You mean not interesting. The author (Yudkowsky) is really a OB who doesn’t understand the situation unfortunately, it’s a bit sad.
He’s writing a whole new theory to be able to make the incorrect choice in the Newcomb problem but avoid the correlation is causality problem in similar situations.
About Yudkowski, he has a whole community and website around him with people who have similar opinions. And this thread is basically a poster showing that among people who OB, we have theists, undergraduates and themselves and among people who TB, we have philosophers and people specialized in decision theory.
The comments are really funny as some would do anything but acknowledge they might be wrong. Some comments:
People suck at predicting their actions. I suspect that in a real-life situation even philosophers would one-box. For example, suppose a two-boxer sees a street magician predict people’s behavior and consistently punish two-boxers (in some suitable version, like a card trick). Odds are, he will one-box, especially if the punishment for correctly predicted two-boxing is harsh enough. It would be an interesting psychological experiment, if someone could get the funding.
More formal logic and philosophy training → a greater chance to over-think it and think explicitly about decision theory (and possibly even have loyalties to particular rigid theories) rather than just doing what gets you more money ? A case of thinking too deeply about the matter just leading a large fraction of people into confusion?
From my cursory understanding of his theory, that’s how I got it too.
Either he is swapping out the near-perfect predictor with a perfect predictor, and that is fundamentally a different scenario.
Or, he isn’t, in which case his theory reduces to Evidential Decision Theory here.
If that cursory understanding is correct, then he is just muddying the discussion with fancy terms, trying to obfuscate the fact that he is not saying anything new.
There’s no trick involved. To repeat the game as specified earlier:
- $1,000 is placed in a blue box
- You are cloned and the clone is asked to choose between either a) the red box or b) the red and the blue box
- The clone is destroyed
- If the clone chose (a) then $1,000,000 is placed in a red box
- You are asked to choose between either a) the red box or b) the red and the blue box
- You win the contents of the box(es) you chose
The clone doesn’t know that he is a clone and you don’t know that you are not a clone. Given that the clone is a complete copy of you (including your brain, and so has all the same knowledge, beliefs, and memories), it will almost certainly reason the same way you reason and choose the same thing you choose. But this might not be guaranteed, hence the clone is a near-perfect, but fallible, predictor.
You just have to accept that your choice will almost certainly be predicted, in which case it is only rational to choose (a), because almost everyone who chooses (a) will walk away with $1,000,000 and almost everyone who chooses (b) will walk away with $1,000.
Yeah, CDT is leading far ahead of EDT among experts. That’s probably because the experts see the full picture.
CDT loses you money in the Newcomb Game V1, if you look across all phases. (CDT wins in Phase 2 though, always)
CDT wins you money in real life, for the most part.
CDT happens to be true.
CDT is therefore rational.