One question I have about decision theory is how people are even coming up with the different decision problems. Some of them only differ in the payoff matrix. But there are other differences like:
the characters in play (Omega, copies of yourself, other players)
randomness (e.g. coin flip in the death in Damascus problem)
the information you get to see (e.g. normal Newcomb vs clear box Newcomb)
I think one way to make progress on understanding this is to try to systematically record existing decision problems, in a machine-readable way (or almost-machine-readable way, like an HTML table). That way, you can start "enumerating" decision problems starting at the simplest, and be able to "locate" the existing problems at some number in your enumeration (I making an analogy with enumeration proofs in logic, but this is probably not useful). Along the way, you will discover other decision problems.
Another way to analogize this is with theorem proving in math. I am thinking especially of "A Fully Automatic Theorem Prover with Human-Style Output" by Ganesalingam and Gowers, and how Gowers likes to say that a lot of undergraduate math proofs can be done on autopilot. Is there some systematic way to understand how people are coming up with decision problems, such that the process can be automated? Is there a way to tell which decision problems are interesting?
Actually, since I am a decision theory noob, it is quite possible someone has already done something like this, so a first step would be to look for existing work.
One question I have about decision theory is how people are even coming up with the different decision problems. Some of them only differ in the payoff matrix. But there are other differences like:
I think one way to make progress on understanding this is to try to systematically record existing decision problems, in a machine-readable way (or almost-machine-readable way, like an HTML table). That way, you can start "enumerating" decision problems starting at the simplest, and be able to "locate" the existing problems at some number in your enumeration (I making an analogy with enumeration proofs in logic, but this is probably not useful). Along the way, you will discover other decision problems.
Another way to analogize this is with theorem proving in math. I am thinking especially of "A Fully Automatic Theorem Prover with Human-Style Output" by Ganesalingam and Gowers, and how Gowers likes to say that a lot of undergraduate math proofs can be done on autopilot. Is there some systematic way to understand how people are coming up with decision problems, such that the process can be automated? Is there a way to tell which decision problems are interesting?
Actually, since I am a decision theory noob, it is quite possible someone has already done something like this, so a first step would be to look for existing work.