T|S One Box or Two Boxes (Newcomb's Paradox)

the premiss that he is accurate in all his predictions isn't prima facie 100% ridiculously implausible

yes, yes it is.

OK, imagine you are Bob.

You predict.

ledge's strategy only works if you assume you can trick Bob into making a false prediction

Bob's prediction doesn't make a jot of difference when you're standing there trying to make up your mind, with the money already in the box/es. Or so a two-boxer would argue.

here's a nice argument:

One can imagine that the Predictor, who is also a truth-teller it turns out, tells your buddy whether or not the million is in the box. What would your buddy, who has all the information about the potential payoffs, recommend that you do? If he had your financial interests at heart, he would always, no matter what the Predictor did, recommend you take two boxes. It would seem completely irrational to go against the advice of your well-informed buddy.

Bluddy ell, now Chris Tarrant is in the room!

also, fwiw

it is a logic puzzle, or at best really watered down game theory

it was apparently orgininally formulated to point out problems with bayesian decision theory, and some people have argued that it's identical with the prisoner's dilemma. iow rather than a fun game for all to play it is more of a thing for philosophers to puzzle over and write lots of papers containing lines like "v(b1) = v(m).p(m/b1)+v(n).p(n/b1) = 990,000".

it's not nearly well defined enough to be that, if we're free to assume that bob is this, that or t'other

All that matters is that Bob is a reliable predictor, it doesn't matter how he does it - it could even be pure luck. That's really the point, in fact - the result is probabilistically dependent on what you pick (because of Bob's history of reliable predictions) - but causally independent (because Bob is only predicting, not affecting the outcome: the money is already in the boxes).

the real question here is, is Bob Newcomb a viking at sleep

Seems strange to give figures for the cash involved, and not for the reliability of Bob's predictions.

I suppose it'd just be a probability equation then, but without it it's just a question akin to 'How long is a piece of string, Bob'?

If the predictor isn't 100% accurate, then this isn't about free will. If the predictor is 100% accurate, then the problem of free will is implied, but not demonstrated by the problem itself. At best, this problem seems to be a good way to stimulate discussion about why the population is split on their decision-making strategies.

yeah it's nothing to do with free will really.

Btw, I think you have to take both boxes. Whatever he predicts, he isn't predicting anything at the moment that you're choosing a box. There's no reverse causality.

― Mordy, Tuesday, May 25, 2010 10:15 PM (Yesterday)

going to disagree with this, the whole question hinges on the idea that your choice creates the status of the boxes. this would be more clear if bob was 100% accurate, but even with the slight margin of error in your formulation, theres no point in taking the super long shot.

still say this isnt a paradox, but rather an uncomfortable counterfactual situation that irritates people because we are wired to be really defensive about free will, and reject reverse causation out of hand. within the structure of bob world, free will (in this case) is suspended, and reverse causation is possible by definition, which removes the paradox.

xposts

going to disagree with this, the whole question hinges on the idea that your choice creates the status of the boxes.

exactly

very happy that my bros with philosophy degrees are backing me up on this, makes me feel like I don't fail at thinking after all

xp only on ilx could you read a post like that from a username like that. i love it.

the whole question hinges on the idea that your choice creates the status of the boxes.

it really doesn't, the problem (as originally formulated) is nothing to do with free will or reverse causality, but about decision theory.

as counterintuitive as it seems on the surface, HI DERE is right about the flux aspect of the money because standard temporal rules are suspended in this case (again, this is even more obvious in the case of the 100% accurate bob). the money in the boxes is indeterminate until you make your choice, which again, i realize is a punch in the directional time part of the brain, but infallible predictors dont exist either so you get what you get.

xpostststsss

Bob (almost) always predicts what you are going to do correctly, regardless of the mental contortions you go through in making your decision. If you decide to take both boxes, he will have predicted that you will take both boxes. If you take one, he will have predicted you will take one. You're betting on him being wrong, something he rarely is based on the setup, ergo by taking both boxes you are betting on the long odds that he has made a mistake, whereas if you take one, you are betting against the log odds that he has made a mistake.

what about your pal who knows what's in the boxes, and always tells you to take both?

going to disagree with this, the whole question hinges on the idea that your choice creates the status of the boxes.

exactly

very happy that my bros with philosophy degrees are backing me up on this, makes me feel like I don't fail at thinking after all

― Marni and Louboutin: coming to Tuesdays this fall on FOX (HI DERE), Wednesday, May 26, 2010 3:31 PM (4 minutes ago) Bookmark

If the predictor is less than 100% accurate, does your choice create the status of the boxes? The prediction creates the status of the boxes, but the predictor could be wrong about your choice. If the predictor is wrong, then your choice had nothing to do with the causality of the status of the boxes.

Here's a more realistic version of the problem, where the answer seems fairly clear (and it's the equivalent of two-boxing). (And it's nothing to do with free will.)

Though it is well established that there is a strong statistical correlation between smoking and lung cancer, it doesn't follow that smoking is a cause of cancer. The statistical association might be due to a common cause (a certain genetic pre-disposition say) of which smoking and lung cancer are independent probable effects. (This is sometimes referred to as "Fisher's smoking hypothesis"). If you have the bad gene you are more likely to get lung cancer than if you don't, and you are also more likely to find that you prefer smoking to abstaining. Suppose you are convinced of the truth of this common cause hypothesis and like to smoke. Does the fact that smoking increases the probability of lung cancer give you a reason not to smoke?

what about your pal who knows what's in the boxes, and always tells you to take both?

he's a dick who has already palmed $99K for himself

I don't think this problem has anything to do with free will. Probability is different than causality, right? The problem seems to be that the probability changes depending on your choice, given a reliable predictor.

If you were Bob, would you predict that the dude "will take 2 boxes"? Or no?

Whatever is the selection, it would always be the same, right?

In which case, Bob is lacking in free will. Just as much as if both boxes were transparent.

For 'selection' read 'prediction' obv.

If Bob was a perfect predictor, then this would be a problem about free will and causality, but he's not, he's a highly reliable predictor, which makes this a problem about probability.

Bob chooses first, and chooses on the basis of the scenario, which never changes, so his prediction never changes.

Mark, if the results of this poll are anything other 100% for one answer, then you are wrong.

Yeah, but the poll is not "You are Bob, what's your prediction"..

Also, Bob 'himself' never changes, whereas all these ILXors have a different viewpoint.

Mark, if I am Bob, then I am not making a choice based on probability theory, I am predicting the choices that others will make based on how they apply probability theory.

Whatever is the selection, it would always be the same, right?

No.

i agree that this works out to be a probability thing because of the not 100%, but the supposed paradox is all about the idea that if the money is already in the boxes, you should always take both boxes vs. choosing one box because your box choosing creates the state of money in the boxes, which means that the paradox attempt here has everything to do with causation and nothing to do with probability.

in other words both boxers are arguing in favor of determinism and one boxers are arguing in favor of reverse causation.

The problem is intended to be about expected-utility principle vs. dominance principle, and it works if it creates a scenario wherein both principles are equally valid. If both principles are equally valid in this scenario and they both lead to different choices, then there is an apparent paradox.

I don't think the mechanism for the reliability of the predictor is important, because the chooser can't know how or why the predictor is reliable. He could be right because he is prescient. He could be right because he has a lot of information, and he's dropping science. He could have been calling heads or tails right, without any special knowledge, every time up until now.

Am I causing the boxes to be in a particular state by making a choice, or is the state of the boxes fixed before I make my decision? There is no evidence contained within the problem to make me choose one way or the other. The question is whether I accept that he is a reliable predictor.

Given that I do accept that he is a reliable predictor, do I apply the expected-utility principle, or the dominance principle? What is the most rational choice?

Am I causing the boxes to be in a particular state by making a choice, or is the state of the boxes fixed before I make my decision? There is no evidence contained within the problem to make me choose one way or the other.

Not as stated in this thread, but it is written into various formulations. Sometimes it's stated that Bob puts the money in the boxes e.g. an hour before you make your choice. Also the version where your friend knows what's in the boxes obviously requires the amount to be pre-determined.

One can make the problem about free will, but that makes it rubbish. If your choice reverse-causes the outcome then determinism doesn't get a look in, you should always two-box.

You mean one-box, right? If you're reverse-causing, then go for just Box B.

woops, yeah.

Found the original articles on the problem and they clearly state that the super-predicting being puts the money in the boxes, then you make your choice.

http://books.google.co.uk/books?id=XOVelCewRMIC&lpg=PA45&ots=D9GeealvwJ&dq=%22Newcomb's%20Problem%20and%20Two%20Principles%20of%20Choice&pg=PA45#v=onepage&q=%22Newcomb's%20Problem%20and%20Two%20Principles%20of%20Choice&f=false

I don't think the mechanism for the reliability of the predictor is important

True enough. It is far more important to know Bob's actual accuracy, which ought to be measurable more exactly than "almost always", than the mechanism of prediction. However, the mechanism is of interest in that it establishes what sort of universe we are inhabiting, and that allows us to judge whether Bob's methods are as likely to be accurate in the matter of boxes filled with cash as they are with other, past actions he has predicted.

BTW, Bob's would appear to be a magical universe, in that there is as yet no imaginable mechanism for predicting my actions in this universe with such near infallibility, without invoking magic, or a diety with magical propoerties. However, if this magical universe is inherently as predictable in as a lawful one, then the mechanism is of no importance, for they will act commensurately. IOW, the magical process must follow laws just as strict and as predictable as physical laws.

I am presuming that, for the purpose of this trial, Bob and I have not already been though this process of choosing hundreds of times already, but that it is novel.

Why is the friend advising you to take both boxes?

Am I causing the boxes to be in a particular state by making a choice, or is the state of the boxes fixed before I make my decision? There is no evidence contained within the problem to make me choose one way or the other.

Not as stated in this thread, but it is written into various formulations. Sometimes it's stated that Bob puts the money in the boxes e.g. an hour before you make your choice. Also the version where your friend knows what's in the boxes obviously requires the amount to be pre-determined.

― literally with cash (ledge), Wednesday, May 26, 2010 5:30 PM (12 minutes ago) Bookmark

If the predictor is prescient, then it doesn't matter when the cash goes in the box--it's still a matter of reverse-causality. I agree that the problem is basically bullshit if the thought-problem is about a being who is infallible.

Why is the friend advising you to take both boxes?

― Marni and Louboutin: coming to Tuesdays this fall on FOX (HI DERE), Wednesday, May 26, 2010 5:54 PM (55 seconds ago) Bookmark

Because he gets a cut of every 100K that doesn't make it into box B.

It's just, if the friend has your best interests at heart and is telling you to take both boxes, then the friend is changing the parameters of the scenario and makes Bob's prediction null and void; the assumption becomes that Bob always puts the $100K into box B and you should always take both because your friend who is, by definition of the problem, not attempting to screw you over knows that the money in the boxes.

actually i dont see any case where the timing of when the cash goes in the box matters. i think what some people are missing is that reverse-causation doesn't imply that the money is actually literally in flux until the instant your decision is made, but functionally thats the only way to view it - the whole thing hangs on the inherent difficulty of prediction wrt linear time. the gap between when the prediction was made/when the cash was put in the box is immaterial really.

xposts

still lolling that the three of us are in hardcore agreement here

Having read the original formulation of the problem by Newcomb (link above), he glosses over the mechanism as irrelevant, and also glosses over what sorts of experience the person asked to make the choice has had with prior predictions, stating only that you know the being (aka Bob):

"has often correctly predicted your choices in the past (and has never so far as you know made an incorrect prediction about your choices) and furthermore this being has often correctly predicted the choices of other people, many of whom are similar to you, in the particular situation to be described below".

I especially like how he plays both sides of the street by saying that you do not know of any predictive errrors in your own case, but that you know (with equal emphasis) that the being's track record in this regard is only "often" correct. Which looks to me like a contradiction: either you know of some errors, or you don't know the being's record is only "often" correct. These can't both be true.

That sort of hand waving makes the problem suck as an exercise in logic or problem solving.

yeah frankly i still feel like the almost always correct (or worse, "often") is just kind of a cheap trick to muddle the obvious one box answer.

To the point of causality, if the predictor is prescient and infallible, and free will still somehow exists, then the state of the cash in box B is "both" until the chooser chooses.

all sorts of xposts

jjjusten, I think the argument is supposed to be that a rational decision-maker would always choose both boxes unless there was a high enough probability that choosing box B will a deliver a big enough reward, and that a statistically reliable predictor is supposed to force a clash between these two rational approaches in decision-making.

I'm not sure how I feel about it, because my decision is supposed to be based on the fact that I accept that there exists a statistically reliable predictor of my behaviorwho is neither prescient nor just waiting to be wrong the next 100 flips of the coin.