Theodicy and the simulation hypothesis, or: The problem of simulator evil — LessWrong