Surely if one applies the hypothetico-deductive method to the hypothesis of an omnipotent, omniscient and benevolent being, one must reject this hypothesis. From the postulate we deduce the non-existence of evil; observation shows of [sic] the existence of evil; hence we must reject the hypothesis. However the matter is not so simple, because as the philosopher of science P. M. M. Duhem taught us, hypotheses come along with auxiliary assumptions which are needed for the rigorous deduction of observational consequences, and a scientist can hang on to his hypothesis by modifying the auxiliary assumptions.
Of course this may be an unplausible thing to do, but this shows how no knock down refutation of a hypothesis is possible. In the case of the hypothesis of an omniscient, omnipotent and benevolent being, the theologian will try to reject one or other of the auxiliary assumptions which (as further analysis might reveal) are needed for the deduction of the non-existence of evil. Still, even so, we may find that his expedients are ad hoc and lack plausibility.
(J. J. C. Smart, "Science and Religion," The Australian Rationalist 3 [November 1973]: 3-7, at 7 [boldface in original])
Note from KBJ: Shorter Smart: If theism implies that there is no evil and there is evil, then theism is false (by modus tollens). But theism alone doesn't imply that there is no evil; theism conjoined with certain other propositions implies that there is no evil. So if there is evil, then either theism is false or one of the other propositions is false. Smart says that the other propositions may be implausible; but plausibility is relative. To a theist, the theistic hypothesis is more plausible than any other propositions with which it is conjoined, so if something in the set has to be rejected, it will be one of the other propositions. Theism, in short, is entrenched.
Keith Burgess-Jackson, J.D., Ph.D. 