I largely agree, but please note that B might be preferable to A and possibly C and D because of considerations of robustness. In the event that one of the two simulations in B is destroyed, you are left with something very close to the isomorphic A rather than nothing at all.
I am assuming p(both simulations are destroyed | B) < p(the simulation is destroyed | A), and also that there is no intrinsic disutility in destroying a simulation provided that other identical simulations exist.
I interpreted it as a question of utility, not expected utility. The expected utility of attempting B is higher because if you fail, at least you get A.
I largely agree, but please note that B might be preferable to A and possibly C and D because of considerations of robustness. In the event that one of the two simulations in B is destroyed, you are left with something very close to the isomorphic A rather than nothing at all.
I am assuming p(both simulations are destroyed | B) < p(the simulation is destroyed | A), and also that there is no intrinsic disutility in destroying a simulation provided that other identical simulations exist.
I interpreted it as a question of utility, not expected utility. The expected utility of attempting B is higher because if you fail, at least you get A.