Rowe's argument is basically:
- E is an evil for which despite serious study we can't see a justifier.
- Therefore, probably, E has no justifier.
- Therefore, theism is false.
Whenever someone gives us an inductive argument involving a particular case, it is appropriate to worry about selection effects in the choice of the case, and incorporate evidence arising from the nature of the selection procedure into the argument. In the case of (1), Rowe did not start with a list of the trillions of evils that happen in the world, and then pick out one, E, at random, and observe that we can't see a justifier for it. There are many evils where people to whom they happened claim to know justifiers, and Rowe picked none of those. Rather, Rowe picked (or imagined) a particular case, E, which was such that we couldn't see a justifier for it despite serious effort.
Thus, once we fill in the story about the selection method, Rowe's argument really is:
- There exists an evil for which despite serious study we can't see a justifier.
- Therefore, probably, there exists an evil that has no justifier.
- Therefore, theism is false.
Now it seems that I've made no progress in rephrasing the argument in this way. You existentially instantiate (4), thereby you get (1), thence you derive (2), and hence you conclude with (5). But this line of reasoning is fallacious in an inductive argument. Here is one way to see this. Let T be the theory that every large galaxy has a central black hole. Suppose we take T to be true. Then, I submit, we have very good reason to think that:
- There is a large galaxy that after serious examination does not look like it has a central black hole.
Consequently, we should be able to existentially instantiate (7) and get:
- G is a large galaxy that after serious examination does not look like it has a central black hole.
- Probably, G does not have a central black hole.
- Therefore, T is false.
Likewise, because we know that there are lots and lots of evils (over a hundred trillion if we do this quick calculation: 10 billion people, each living about half a century, each experiencing about one evil a day), we have good reason to expect that even if every evil has a justifier, some of the justifiers are going to be more obvious and some are going to be less obvious. And among those that are less obvious, there will be ones that are (roughly) the least obvious--and those are likely to be quite unobvious indeed. Thus, we have very good reason to expect that (4) is true even if every evil has a justifier. And hence (4) is not significant evidence against the claim that every evil has a justifier.
To put it briefly: Theories that apply to many cases are likely to have both weak and strong "anomalies". C is a strong anomaly for a theory T if independently of the evidence for T, C looks like a counterinstance to T. C is a weak anomaly for T if independently of the evidence for T, it's not the case that C looks like an instance of T. The mere fact that a general theory has an anomaly is not in general significant evidence against the theory. Scientific practice reflects this.
It is pseudoscience that latches onto unrepresentative anomalies. Think of pseudoscientific criticisms of evolutionary theory (I am not claiming that all criticisms of evolutionary theory are pseudoscientific). "Nobody has come up with an evolutionary explanation of biological feature F despite serious work. Therefore, F has no evolutionary explanation. Therefore, evolutionary theory is false." This is a bad argument, because armchair biology would lead us to think that evolutionary explanations will sometimes be hard to find, even if all biological features have them.
The fallacy in question is an interesting one. There are two ways of trying to analyze it. One is that existential instantiation is problematic in inductive arguments. That's an intriguing suggestion. The second is simply that the argument confuses what is we see independently of the evidence for T and what we see given the evidence for T. There are going to be cases where our only reason for thinking that they fit with T is the general evidence for T.
Anomalies are to be expected and the mere fact that a theory has an anomaly does not make us question it. Indeed, the lack of an anomaly can make one suspicious (if all of an experimenter's data looks too close to some curve, one is worried about fraud). Of course, anomalies may pile up. Or they may turn out to be representative. If computer simulation shows that the black-hole-center theory predicts that only 5% of galaxies would look like they don't have a central black hole, but 10% look like that, then that can be strong evidence against the theory. Interestingly, though, if the theory predicts that 5% would look like they don't have a central black hole, but only 2% look like that, that's also a problem for the theory!
Likewise, Rowe's argument from evil can be replaced by new arguments, such as (a) that the percentage of apparently unjustified evils is greater than that which would be predicted by the theory that all evils have justifiers, or (b) that the number of evils is greater than theism predicts, or (c) that God couldn't permit there to be a single evil for which a justifier is not evident. But each of these is a significantly different argument from Rowe's original argument: (a) requires serious statistical investigation; (b) is messy; and (c) is deductive (I am grateful to Adam Pelser for this point), and hence needs only a defense to be refuted.
Observe that this response is not a sceptical theist response. It is not sceptical: on the contrary, it is thoughts like these that keep us from been sceptics about scientific theories. And the response is not based on any controversial theses about the realm of value. Rowe simply commits a fallacy of non-deductive reasoning, the fallacy of taking anomalies too much to heart. (Perhaps this is a fallacy that occurs elsewhere in philosophy.)
Notice also that while one might, and probably should, worry that (1) is apt to confuse absence of seeing with seeing of absence, even if one grants (1) with the stronger "seeing of absence" reading (which corresponds to "strong anomaly"), the point remains.
I do not dispute that the existence of a single anomaly is incremental disconfirmation of the theory. But the existence of a single anomaly--or even of a moderately large number of them (if 0.1% of the hundred trillion evils are likely to be anomalous--surely a not unreasonable estimate--the numbers of anomalous evils will be large)--does not significantly lower the probability of a theory. Such incremental disconfirmation should not worry theists. Rowe's argument is, I think, over.