There are two ways to object to atheistic inferences to the best explanation. One is to criticize the reasons these arguments offer for thinking some atheistic hypothesis has more theoretical virtue than the theistic hypothesis. The other, which goes much deeper, is to doubt that inference to the best explanation (henceforth, “IBE”) is a reliable type of inference at all. I have seen theists take both approaches. Today I’m going to show that the second approach is not as successful as theists might hope.
Two of the most popular and, in my opinion, powerful objections to IBE as a form of inference are the argument of the bad lot and the argument from unconceived alternatives. These arguments are very similar. The first was developed by philosophy of science rock star Bas van Fraassen, and it goes like this: Even if the explanation one argues to be the best is better than all the other explanations already on the table, that doesn’t mean it’s better than all conceivable or possible (or available, in one sense) explanations. It might be the best of all the explanations that have been considered, but still be much worse than the best possible explanation of these particular data. In fact, it might be so much worse that it’s a bad explanation. All the explanations we’ve considered might be bad, making the best explanation we’ve come up with nothing more than the best of a bad lot. But if we’re not in a position to say with confidence that our best explanations aren’t bad, then even if theoretical virtues are good guides to the truth, we can’t say our best explanations are probably (approximately) true. At best we can say they’re way more likely to be true than all the rival theories that have been proposed.
Philosophy YouTuber Kane B presents Kyle Stanford’s problem of unconceived alternatives as follows:
(P1) The historical record reveals that there were equally good alternatives to the accepted theories of the past, but that past scientists failed to conceive of.
(C1) So, probably there are equally good alternatives to current theories that current scientists fail to conceive of.
(C2) We should not believe that currently accepted theories are true.
As you can see, this is extremely similar to the argument of the bad lot. The primary difference is that it appeals to historical evidence in support of the claim that it’s likely we’re working with either a relatively bad lot, or a lot that omits alternatives worthy of equal degrees of belief. If we were to conceive of those alternatives, we would at the very least suspend our judgment as to which are true: those theories or the currently accepted theories.
Never mind that Stanford’s argument focuses on the history of scientific theory choice rather than the history of theory choice more broadly (which includes philosophical theory choice, such as choosing between theistic and naturalistic theories). It’s plausible enough that the history of all theory choice looks about the same.
Here’s why these particular doubts about IBE shouldn’t be of much comfort to the theist. Even if these arguments against IBE succeed, they don’t undermine arguments that theism is a failed theory. As long as theism is a much worse theory than, say, naturalism, then for atheological purposes it matters very little whether naturalism is just the best of a bad lot. What the theist needs is to argue against our ability to weed out false theories, not just our ability to identify true theories. As Paul Draper writes on p. 15 of “Where Skeptical Theism Fails, Skeptical Atheism Prevails”:
To show, for example, that the falsity of theism is highly probable, it need only be shown that theism is many times less probable than one specific alternative theory. This suffices because, if one statement is many times more probable than another, then even though it doesn’t follow that the first is probably true, it does follow that the second is very probably false. This is why I have in the past constructed what I like to call ‘Humean’ arguments from evil against theism.
(For those who are unfamiliar, here’s an in-depth summary of Draper’s celebrated Humean argument from pain and pleasure.)
Of course, if we can’t show naturalism is probably true, that’s an issue for the atheist philosophers, like Graham Oppy and Jeffery Jay Lowder, who are invested in the project of establishing naturalism via abduction (another word for IBE). But it’s not a big issue for defenders of generic atheism, and it’s hardly a win for theists.
Moreover, if the theist comes to doubt IBE’s ability to identify true theories, they should no longer be persuaded by IBE-based arguments for God’s existence, like those presented by Richard Swinburne, David Baggett & Jerry Walls, and William Lane Craig.
Even if these arguments against IBE succeed, they don’t undermine arguments that theism is a failed theory.
Objection: Ben, you’ve just conceded that considerations of theoretical virtue and vice may be no good at revealing what’s true. But if such considerations successfully reveal that theism is false, they thereby reveal that it’s true that theism is false. How then can you claim that these considerations are good at exposing theories as false? Make up your mind, man.
Reply: To argue abductively that theism is false is indeed to argue for the truth of the proposition that theism is false. But it’s not to argue that some explanation is the best, and thus probably true. For the proposition that theism is false really doesn’t explain anything, nor is it meant to. It only tells us that something doesn’t exist, which typically doesn’t help us account for what does exist. The abductive atheological arguments instead proceed by contending that certain positive alternatives to theism, like naturalism, are much better explanations than theism. These alternatives are positive in that they do say something exists, such as the physical world — something that can play a role in explaining the data. So abductive atheologians don’t make the faulty inference that, just because the negation of theism is the best explanation we have of the data, it’s not a bad explanation. They don’t say the negation of theism is an explanation in the first place, let alone a good one.
Now, I wholeheartedly acknowledge that there are other ways to object to the reliability of IBE, some of which seek to undercut inferences to theories’ falsity. For example, there are worries about how to assess the relative probabilities of theories A and B when A is better in one way, say simpler, and B is better in another, say more consistent with our background knowledge. And there are worries about which, if any, of the theoretical virtues and vices count as evidence that a theory is more or less probably true than another. Why, for instance, should we think the world conspires to make simple theories true more often than complex theories? Those worries will have to be addressed independently. But I wanted to emphasize that abductive atheologians like Draper have little to fear from bad lots and unconceived alternatives.
“To argue abductively that theism is false is indeed to argue for the truth of the proposition that theism is false. But it’s not to argue that some explanation is the best, and thus probably true.”
I was thinking about the historical episode between Copernicus’s heliocentrism and Ptolemy’s geocentrism. Heliocentrism was falsified when no stellar parallax was observed. This apparent lack of stellar parallax was better explained by geocentrism at the time.
But I believe you already considered this:
“For example, there are worries about how to assess the relative probabilities of theories A and B when A is better in one way, say simpler, and B is better in another, say more consistent with our background knowledge.”
After reading the above I was left confused as to why you still seem to think that abductively arguing some explanation is false is in some way reliable.
Thanks, this is the perfect historical example to consider. I have to admit I’m not too familiar with the details, but given how you’ve described the case, it looks as though what happened was a fairly reasonable rejection of heliocentrism. I take it that your concern, then, is that this demonstrates what we can by abductive standards reasonably deem false may turn out to be true. This was true of heliocentrism, so it might also be true of theism.
As far as I can tell, the worries I mentioned in your quote don’t come up in the heliocentrism case. It’s not that heliocentrism and geocentrism were ever such that one was better in one explanatory respect and the other better in another, and the relative weights that we should’ve given to these distinct virtues were uncertain. Rather, geocentrism seemed better all around given our limited knowledge at the time, and that was why we opted for geocentrism. And arguably it even *was* the better explanation, despite being false! But since in our armchairs we aren’t limited by historical contingency, we may, if we wish, conceive of a case just like the historical clash between geocentrism and heliocentrism but with a trade-off of theoretical virtues. I won’t do that in this comment.
Of course, it’s not good enough in this case to say, “Oh, we just shouldn’t accept abductive inferences to the truth of explanations. That was the mistake we made. We should’ve admitted that geocentrism could be false despite being the best theory, because the unconceived alternatives could be much better.” That would make up for one mistake, but it wouldn’t make up for the mistaken conclusion that heliocentrism is false, which those living back then would’ve reached had they taken the approach I propose in the post.
My first reply, then, is that one historical example of my approach’s going wrong doesn’t suffice to show that abduction to falsity is unreliable. Reliability only requires the approach’s getting things right the vast majority of the time. Secondly, it’s plausible to regard the historical rejection of heliocentrism as having been rational, even though it was a factual error and in retrospect might’ve been based on an unreliable form of inference. Having said that, there’s a concern that, once we realize a form of inference is unreliable in that it fails to track the truth, it’s no longer rational to use it (or if it’s still rational, what good is rationality?). Suppose that’s right. That brings us to my third reply, which is that the way you describe the case may either be misleading or underspecified, for all I know. It may be misleading in that, if we had reasoned more carefully, we would’ve seen even back then that heliocentrism was not a (significantly) worse explanation than geocentrism. One reason to suspect that’s true is that you only cite one problem, the apparent lack of stellar parallax, as having falsified heliocentrism. If that was the only basis for rejecting heliocentrism, perhaps other factors relevant to the relative probabilities of geocentrism and heliocentrism were irresponsibly dismissed. Or your description may be underspecified in that there was some relevant information available to people back then which, while not being irresponsibly dismissed, wasn’t even taken into consideration. Had it been taken into consideration, reasonable people would’ve decided heliocentrism was the better explanation.
“My first reply, then, is that one historical example of my approach’s going wrong doesn’t suffice to show that abduction to falsity is unreliable. Reliability only requires the approach’s getting things right the vast majority of the time. ”
Well, then I think we can at least agree that abduction to falsity can sometimes lead to false beliefs.
“Secondly, it’s plausible to regard the historical rejection of heliocentrism as having been rational, even though it was a factual error and in retrospect might’ve been based on an unreliable form of inference.”
Could it be that this unreliable form of inference was abductive reasoning? Geocentrism was the best explanation for the time.
” If that was the only basis for rejecting heliocentrism, perhaps other factors relevant to the relative probabilities of geocentrism and heliocentrism were irresponsibly dismissed. ”
No there were other reasons but that was the strongest evidence against heliocentrism. The point is that if heliocentrism was true then, they reasoned, they should of been able to observe a shift in the stars. They didn’t.
“… Had it been taken into consideration, reasonable people would’ve decided heliocentrism was the better explanation.”
If I remember correctly the error was in the background “knowledge”. The universe was believed to be a lot smaller than we later believed. Observing a shift in the stars would of been more obvious in a small universe.
If this historical example interests you may be interested in reading Thomas Kuhn.