which encodes the creator’s arbitrary “logical degrees of caring”, just like its regular prior encodes the creator’s arbitrary degrees of caring over physics
(Has there been any work on moving from an arbitrary point of timestamping to something that obeys something like dynamic consistency requirements? One could think up decision problems where two UDTs would have coordinated except their moment of caring-encoding was arbitrarily single-pointed in spacetime, then try to use such examples to motivate generalized principles or notions of consistency. That’s a different way of advancing UDT that seems somewhat orthogonal to the focus on self-reference and logical uncertainty. (The intuition being, of course, that arbitrariness is inelegant, and if you see it that’s a sign you need to go meta.) Maybe Nesov’s desire to focus more on processes and pieces rather than agents will naturally tend in this direction.)
(Has there been any work on moving from an arbitrary point of timestamping to something that obeys something like dynamic consistency requirements? One could think up decision problems where two UDTs would have coordinated except their moment of caring-encoding was arbitrarily single-pointed in spacetime, then try to use such examples to motivate generalized principles or notions of consistency. That’s a different way of advancing UDT that seems somewhat orthogonal to the focus on self-reference and logical uncertainty. (The intuition being, of course, that arbitrariness is inelegant, and if you see it that’s a sign you need to go meta.) Maybe Nesov’s desire to focus more on processes and pieces rather than agents will naturally tend in this direction.)
Um, aren’t timestamped preferences already dynamically consistent?