withallyourmind.net

In my last post, I tried to give a (very) brief summary of William Dembski’s explanatory filter, which he claims is a slightly more formalized method of the way we commonsensically infer design around us all the time. Dembski also claims that his explanatory filter is already used by many scientists in respectable fields such as forensics, computer science, SETI, and archaeology. The hope, I think, is that if the explanatory filter works, and is already used to detect design in other sciences, then it should also be tolerated as a method for inferring design in other sciences, principally biology and cosmology. Although I think that the explanatory filter is on to something and accurately depicts the way we sometimes come to infer design, I think it has (at least) a couple flaws that make me hesitant to place much faith in it as a big step forward for the Intelligent Design Movement.

The first problem is with the very first step, in that it sets up a false dilemma between design and law. Something that intelligent design advocates have rightly stressed for a long time, especially when answering charges of committing some sort of god-of-the-gaps fallacy, is that design inferences can be justified even in cases where there is no direct divine intervention. That is, even if some event is completely explicable by reference to natural law and initial conditions, it may still be rightly attributed to design. For example (I think this is adapted from Ratzsch’s philosophy of science book), if we someday land on Pluto and find, when we arrive, a series of craters caused by various meteors that, when seen from afar, spell out “Welcome to Pluto,” then even if we can trace the paths of each one of the meteors and their causes and find out they are explainable by law all the way back to the big bang, we would still be justified in inferring some sort of design. This is an extreme example, but I think it proves the point.

Now, it may be said in reply that the explanatory filter is only useful for detecting design where some sort of extra-natural intervention occurred. For now, I will concede the point, although I will have something to say about it in a future post. Nevertheless, even if this were the case, I think that the explanatory filter would already prove less useful than it should be, because, as I stated before, there could be cases where design is justifiably inferred even when an event is explicable by natural law, and we should want to draw that conclusion in these cases. Furthermore, it could turn out that most, or the most important, cases where we can detect design in nature turn out to also be explainable by reference to law.

My other main concern is the problem of specification. How is it that we can tell whether some complex pattern is specified or not? Other than just giving some examples where we can intuitively tell that one pattern is specified and another is not, it’s hard to spell out exactly what specification is. The general idea, as I said in the previous post, is that the pattern is separable from its concrete context. Another way to describe a pattern as specified is to say that we could independently produce the same pattern by using our side information. So in the case of the scrabble letter example, given our knowledge of the English language and how words are formed with letters, we recognize the the event of the scrabble tiles coming to form words as the work of design because we could come up with that same pattern independently of the scrabble tiles being thrown and could specify that independently derived pattern as an example of something that an intelligent source would bring about if it were to arrange scrabble tiles.

Hopefully that description was understandable. If not, I’m sure it is partly, if not mostly, due to my lack of clear exposition. However, I think it is also partly because there is something slightly problematic about the notion of specification that makes it so difficult to expound. Specifically, I think it is circular and undercuts the whole purpose of the explanatory filter. To see this, think of some long combination of zeros and ones. Some obvious combinations that would be good candidates for examples of specified complexity would be all zeros or all ones or a pattern that lists the ascending prime numbers in binary code. However, depending on our side information, we could independently specify any combination of zeros and ones. Or, with some sort of algorithm, we could independently specify every possible combination. Nevertheless, we wouldn’t then think that every combination therefore exhibited specified complexity and therefore was designed. But why not? Well, because although we could specify all these combinations, they wouldn’t be specified in the right way, or with the right side information. But, and here’s the main point, what could it possibly mean to be specified in the right way? Maybe I haven’t thought through this all the way, but the only possible informative answer to this question that I can think of is to say that the pattern has to be the type of pattern that we know or believe would be designed. But if this is the case, then we come to believe something exhibits specified complexity by believing that it exhibits design. But if looking for specified complexity is the way to figure out if something is designed, then the explanatory filter becomes reduced to looking for design by deciding whether one thinks something is designed. This is obviously circular.

One thing that I conclude from this, and that I think could be realized independently, is that we don’t generally recognize design by any sort of formal thought process. Instead, we come to have design beliefs in the same way we come to have perceptual beliefs - not by any sort of reasoning based on prior beliefs, but in the basic way of simply coming-to-believe. Now, this doesn’t mean that inferences had in this way are any less justified, or that I think they have no place at all in science. It just means that they may not fit into science the same way that some people in the intelligent design movement want them to.

To conclude on Dembski’s explanatory filter, I think it is flawed because it doesn’t allow us to detect possibly real design that arose due to natural laws, because it’s account of design-detection is at least partly circular, and because it doesn’t accurately capture how we actually make at least the majority of our justified design beliefs.