The initial intuition for one boxing is that you see a bunch of one-boxers walking around with a million dollars and two-boxers walking around with a thousand dollars and looking silly. But it wouldn’t seem silly if the one-boxers only have $1000.01. In this scenario I’d two-box rather than risk the $1000. And it isn’t just minimizing regret. And it isn’t due to not fully accepting the hypothetical (like questioning the setup or reliability of the predictor). I’m asking about uncertainty about the DT itself.
Would any “one-boxers” here always one-box no matter how tiny, tiny the utility difference is? Alternatively, if you change based on the amounts, do you accept having some arbitrary parameter in your DT for something like DT uncertainty? Something else?
The initial intuition for one boxing is that you see a bunch of one-boxers walking around with a million dollars and two-boxers walking around with a thousand dollars and looking silly. But it wouldn’t seem silly if the one-boxers only have $1000.01. In this scenario I’d two-box rather than risk the $1000. And it isn’t just minimizing regret. And it isn’t due to not fully accepting the hypothetical (like questioning the setup or reliability of the predictor). I’m asking about uncertainty about the DT itself.
Would any “one-boxers” here always one-box no matter how tiny, tiny the utility difference is? Alternatively, if you change based on the amounts, do you accept having some arbitrary parameter in your DT for something like DT uncertainty? Something else?