Tuesday, May 18, 2010

Successful Theistic Explanations (Part 1) - Testability

I previously covered chapter 5 of Gregory Dawes's book Theism and Explanation. I am now going to cover chapter 7 of the same book.

Before we get into the meat of this chapter, a quick summary of Dawes's basic thesis is in order. Dawes argues throughout his book that theism can be a genuine explanatory hypothesis. To be precise, it can be an abductive intentional explanation. This means that it will explain events and states of affairs by appealing to divine intentions. Any posited intentions will be constrained by the rationality and optimality principles. These were covered when discussing chapter 5.

Up to this point in the book, Dawes feels he has established the explanatory potential of theism. The final consideration is whether theism can be a successful explanation. This will require an assessment of a potential theistic explanation against a list of explanatory virtues.

In chapter 7, Dawes assesses potential theistic explanations against six explanatory virtues: (i) testability; (ii) consistency with background knowledge; (iii) past explanatory success; (iv) simplicity; (v) ontological economy and (vi) informativeness.

Some of these are more important than others. In this first post, we will deal with the testability of theistic explanations. There is quite a lot of ground to cover here so I hope you are sitting comfortably.

1. Testability and Corroboration
Testability is often singled out as the hallmark of scientific explanations. This may seem to make its application to theism questionable. We'll get to that in a moment.

The first thing to note is that an explanation is testable only if it makes predictions about facts other than those it purports to explain. Another way of putting this is to say that a testable explanation will be able to exclude at least one possible state of affairs. This avoids the "explains everything, therefore nothing" objection.

An example might be my claim that my car won't start because the engine is flooded. This explanation is testable because it predicts that if the engine were not flooded, the car would start. It thus excludes the "not flooded" state of affairs from its scope.

The second thing to note is that mere testability is not enough. An explanation must actually pass the test in order to merit our consideration. When an explanation passes a test we refer to it as being corroborated.

2. Are Theistic Explanations Testable?
The problem facing theistic explanations is that they tend to fall into the "explains everything, therefore nothing"-category. After all, theists tend to believe that everything is ultimately attributable to god (except, perhaps, the freely willed actions of humans), even if the road to that attribution is unclear.

Nonetheless, Dawes argues that theistic explanations could be testable if they were more specific. They would need to identify some divine goal and show how a particular state of affairs served as a means to that goal. They could then be corroborated by showing how this divine goal explains other states of affairs.

The example Dawes uses to make this point is the 2004 Indian Ocean Tsunami. In the aftermath of this incident, at least some religious commentators attributed it to God's desire to stamp out sexual immorality. These commentators are making a specific claim about God's goals, and this claim can be tested. How? By seeing whether other natural disasters target regions known for breaching the sex code.

Of course, as Dawes notes, no reputable religious philosopher is willing to do this because the explanatory merit of the specific goals tend to unravel. For instance, Richard Swinburne has offered probably the most sophisticated defence of merits of theistic explanation. But in doing so he strategically avoids specifying what God's intentions are and how creating this world helps him to realise those intentions.

In sum, theistic explanations in their most sophisticated form tend to be untestable, and in their least sophisticated form they tend to be uncorroborated.

3. Predictions and Retrodictions
After critiquing Swinburne's lack of specificity, Dawes takes a more detailed look at how theistic explanations could make predictions.

The strict understanding of predictions comes from Karl Popper. He argued that in order to make a prediction, a scientific theory must predict wholly new events that can subsequently be verified (or, rather, falsified) by observation.

This Popperian view is unreasonable given the history of scientific practice. For example, one factor counting in favour of Einsteinian relativity over Newtonian mechanics was its ability to account for the long-known peculiarities in the orbit of Mercury. This was not a novel prediction, but it seemed reasonable to accept it as corroboration nonetheless.

But when rejecting the restrictiveness of Popper's view, we must be careful not fall prey to correlative vice of permissiveness. We do this, according to Dawes, by making sure we consider alternative explanations during the corroboration process.

This is referred to as the historical approach. Why? Because the alternative explanations are usually predecessors to the proposed explanation, as in the Einstein-Newton example given above. The new explanation is corroborated if it accounts for something that the old explanation does not account for (and should have been able to account for).

This could be an issue when looking at theistic explanations because there is often no competing explanation. For example, Richard Swinburne claims that God can explain the laws of nature. There is no predecessor theory that attempts to do the same. Does theism then win the explanatory battle because it is the only competitor?

To answer that question, we must take a detour into some of Elliot Sober's arguments against Swinburne.

4. The Solitary Explanatory Hypothesis
Sober objects to Swinburne's thesis on the grounds that it is an untestable, solitary explanatory hypothesis. Sober, like Swinburne, couches his arguments in terms of confirmation theory so his specific claim is that meaningful probabilities cannot be assigned to the solitary hypothesis.

We can see this if we consider Swinburne's argument more carefully. Swinburne is claiming that the probability of there being laws of nature (O) is high on the hypothesis of theism (T) and low in the absence of an explanation. This gives us the following:

  • Pr (O | T) > Pr (O)
Sober's point is that Swinburne has no right to assign a low probability to O and a high probability to O|T. And the main reason that he has no right to do so is that there is no competitor theory.

If Sober is right, then Swinburne's philosophical programme is severely damaged. But is he right? Dawes offers two reasons for rejecting Sober's argument.

First, it might be possible for the theistic explanation to be compared with a "null hypothesis". This is routinely done in scientific investigations. For example, when testing the effectiveness of a drug. The proponents of the drug predict that it will lead to measurable differences between two groups of experimental subjects (those who are given the drug and those who are not). The null hypothesis predicts that any differences between the groups will be attributable to chance.

Something similar could be done in Swinburne's case. In this instance, the null hypothesis would be that the laws of nature got to be the way they are by chance. This would involve defining the relevant probability space (i.e. the range of values or forms that the laws of nature could take). Something that could be difficult to do.

The second response to Sober is more interesting. Sober's argument is, surprisingly, being generous to Swinburne. It accepts the vague form of theism that Swinburne adopts. Dawes would argue that Swinburne should step up to the plate and posit a specific divine intention as the explanation for the laws of nature.

If Swinburne does this, then there will usually be some clear-cut implications that follow. In other words, additional facts that would be explained by that same intention. We could then go out and see whether those facts actually obtain. In this manner, the solitary hypothesis could become testable.

It is, perhaps, telling that Swinburne does not do this.

Okay, that's it for now. In the next part of this series we will consider how well theistic explanations fare when measured against the explanatory virtues of consistency with background knowledge and past explanatory success.


  1. I am also interested in the study of very early christianity. I am currently reading, Walter Bauer's _Orthodoxy and Heresy in Earliest Christianity_. I have studied the topic for almost 20 years now. I have read a number of other of the great works on this topic like Strauss, Harnack, Schweitzer, two additional Bauers (FC & Bruno), and many many others, but this is my current read.

    I am always interested in meeting others that are interested in the study of earliest christianity to have ongoing conversations and share reading lists, etc... you can contact me by email at RichGriese@gmail.com

    Do you have specific aspect of the study that interest you, that you might be interested in discussing, and perhaps having on going discussions on the topic in general? Feel free to email me to talk about it.

    My main interest is the very earliest period. Perhaps from the modified Messiah idea that may have begun around the time of the Maccabean revolt through the beginnings of christianity itself, till the Council of Nicea in 325CE, and perhaps a few years after that as some of the results of that council took effect.

    On my main site I don't post my christian history posts, since most people have no interest in christian history, so I try to make my main site more of a general purpose one. But I do post on christian history topics, and collect resources and links that others interested in the subject may find useful. I am happy to share these with folks that are interested in that kind of thing.


  2. I would note that J.F. Tipler has the distinction of being the only person I know of that offers a testable theology. While his predictions can be disjointed from his theology and turned into mere physics, it is at least an effort and inclination to be respected, in my opinion. Not that I think Tipler will cease to be religious if his physical theories will be disproved, but hey.

    It goes without saying that in the process Tipler makes very specific claims about what god is, allowing him to derive these specific predictions.


  3. I haven't read any Tipler.

    Is he like Gerald (or Gerard, can't remember which) Schroeder? I found his stuff to be extremely ad hoc.

  4. John D,
    must Swinburne take into account both the null hypothesis approach as well as the positing of specific divine intentions in order to make his theistic hypothesis testable or is it enough to consider just one of them? I say this because it seems to me that he might pull it off with the null hypothesis. I mean yeah, it might be difficult to define "the range of values or forms that the laws of nature could take", but I don't think that's going to stop him from coming up with some answer.

  5. Bogdan,

    I would say both. One of the tests of a good explanation is independent confirmation. That is, its ability to explain facts outside of those it is originally introduced to explain. Specifying the intentions helps to do this by giving greater empirical content to the explanation.

    If the posited explanation was simply "God intended to create the universe with the laws that we observe" then we would have gained no understanding from the explanation. This is because the goal of the intention is limited to the fact that we want explained.

    If the posited intention was "God created the universe with the laws that it has because these laws served his goals X, Y and Z" then we have something with real bite: we can point to other facts that might also be explained by the same intention.

    As to whether Swinburne could pull it off with just the null hypothesis, here we begin to cut into the fine-tuning argument. I personally think there are serious problems with calculating or determining the relevant probabilities that go into this argument. So I am not sure he could do it, not that that will stop him or anybody else from making the argument.

    I am hoping to cover some of the problems with FT arguments on this blog, at some stage. In the meantime, I would recommend listening to Luke's interview with Neil Mansun over at commonsenseatheism, as well as reading some of the paper's linked in the same post, below.