One thing that I disagree with in your analysis is that one boxers care about dispositions whereas two boxers care about decisions. I think it’s actually the opposite.
I myself am a fairly confident one boxer who feels the force of two boxing reasoning. I work at a trading firm and found that my firm seems to attract people who think like two boxers. I found a highly intelligent coworker (who seems to be even more intelligent than I am) who is a staunch two-boxer. He seems to understand every point I have brought up, and he brings up fair points.
During our debate, it became apparent that he assumed that, since you can’t change who you were at scan time, and since the prediction (and hence money in the opaque box) follows from who you were at scan time, that the opaque box has already been filled or not, so you might as well make an extra thousand bucks since you can no longer take any action to affect the contents of the opaque box in any way.
I countered by saying that the predictor is not predicting your disposition. It is predicting your output to the decision problem. Imagine that you have to press one of two buttons to make your decision to Newcomb’s problem (and that pressing both simultaneously corresponds to two-boxing). Also imagine that the population of people who get predicted suffer from involuntary muscle spasms that could cause them to press one or more buttons that they did not intend to press. Say that even in cases where the button press stemmed from an involuntary muscle spasm, the predictor is still overwhelmingly accurate. (This could definitely happen if the proclivity and direction of the muscle spasm is foreseeable at prediction time.)
In this world of involuntary decisions, the predictor is still predicting the decision itself. You could think like a two boxer all you want, but if you somehow involuntarily one box anyway, the predictor would have predicted that, and you would reap the rewards of a one boxer despite having the disposition of a two boxer.
We can extend this muscle spasm argument to the normal world, too. The predictor is predicting what you actually decide. Nothing less, nothing more. You could have the disposition of a two boxer all you want, but if you somehow one boxed anyway, you would be a one boxer (by definition).
You may not be able to control your disposition at scan time, but what actually matters is what you actually decide. The predictor’s only goal is to predict your decision.
The two boxer has to be convinced that your decision follows from your disposition in order to conclude that it’s worthless to one box. The one boxer says that it doesn’t matter what you think, if you one box, you reap the rewards of a one boxer, regardless of how you got there.
One thing that I disagree with in your analysis is that one boxers care about dispositions whereas two boxers care about decisions. I think it’s actually the opposite.
I myself am a fairly confident one boxer who feels the force of two boxing reasoning. I work at a trading firm and found that my firm seems to attract people who think like two boxers. I found a highly intelligent coworker (who seems to be even more intelligent than I am) who is a staunch two-boxer. He seems to understand every point I have brought up, and he brings up fair points.
During our debate, it became apparent that he assumed that, since you can’t change who you were at scan time, and since the prediction (and hence money in the opaque box) follows from who you were at scan time, that the opaque box has already been filled or not, so you might as well make an extra thousand bucks since you can no longer take any action to affect the contents of the opaque box in any way.
I countered by saying that the predictor is not predicting your disposition. It is predicting your output to the decision problem. Imagine that you have to press one of two buttons to make your decision to Newcomb’s problem (and that pressing both simultaneously corresponds to two-boxing). Also imagine that the population of people who get predicted suffer from involuntary muscle spasms that could cause them to press one or more buttons that they did not intend to press. Say that even in cases where the button press stemmed from an involuntary muscle spasm, the predictor is still overwhelmingly accurate. (This could definitely happen if the proclivity and direction of the muscle spasm is foreseeable at prediction time.)
In this world of involuntary decisions, the predictor is still predicting the decision itself. You could think like a two boxer all you want, but if you somehow involuntarily one box anyway, the predictor would have predicted that, and you would reap the rewards of a one boxer despite having the disposition of a two boxer.
We can extend this muscle spasm argument to the normal world, too. The predictor is predicting what you actually decide. Nothing less, nothing more. You could have the disposition of a two boxer all you want, but if you somehow one boxed anyway, you would be a one boxer (by definition).
You may not be able to control your disposition at scan time, but what actually matters is what you actually decide. The predictor’s only goal is to predict your decision.
The two boxer has to be convinced that your decision follows from your disposition in order to conclude that it’s worthless to one box. The one boxer says that it doesn’t matter what you think, if you one box, you reap the rewards of a one boxer, regardless of how you got there.