Thursday, May 11, 2006

How to Really Detect Design

While I'm picking on Dembski's blog, I figured I'd critique another entry there, this one written by Salvador Cordova. Sal is a congenial fellow, most of the time at least, so I'll spare him the snark and get right to the point. He writes:

The corporation known as Genetic-ID (ID as in IDentification, not ID as in Intelligent Design) is able to distinguish a Genetically Modified Organism (GMO) from a "“naturally occurring" organism. At they claim:

Genetic ID can reliably detect ALL commercialized genetically modified organisms.

I claim that detecting man-made artifacts (like a GMO) is a valid instance of applying the Explanatory Filter.

To make a long story short, this company attempts to make a buck by detecting whether or not something is a genetically modified organism (GMO). In other words, whether it was "designed" in a sense by humans as opposed to being "natural". According to the company's website, they mostly use their techniques for detecting GMO food, to make sure the food adheres to certain regulations and whatnot. Some countries have banned GMO food (for reasons I think are mostly foolish, but that's another story), so there is a market for such testing.

Cordova's argument is that since this company is detecting the presence or absence of "design", then they are using Bill Dembski's Explanatory Filter. We critics have been calling ID scientifically useless all this time, but we are wrong! ID really is useful. In fact it's being used by this very company!

Well, not quite. There are a couple of serious problems with Sal's argument. First and foremost, we have to understand what the Explanatory Filter (EF) is and what Dembski says it tells us. For the purpose of discussion, we'll grant that the EF actually works as advertised. (If you want a technical discussion of its faults go here.) You can see from the schematic on the left, the EF basically works like this: Everything is to be tossed into one of three bins, Regularity, Chance, and Design. If a specified event (we'll accept for the time being that "specified" is a meaningful concept) has a high probability, then we say it occurs with law-like regularity. If the event has medium probability, then we say it was by chance. But if it has very low probability, then we assign it to design. The filter has ever decreasing inclusiveness as we go on to the next step; in other words, a high probability event may be caused by design, but a low probability could not have been caused by chance.

First of all, this is not what the people at Genetic-ID are doing at all. They are not deciding that a sequence of an organism's DNA is "designed" by figuring that it was just too improbable to have come about by chance or by some law-like process. What they are doing instead is taking the known hallmarks of a GMO and testing for their presence. They explain this on their website:
PCR [polymerase chain reaction] is the gold standard for GMO testing. GMO tests based on PCR are used worldwide to verify contracts and regulatory compliance.

PCR works by massively replicating certain telltale sequences in the DNA. Each variety or genetically modified organism has key sequences that can be used to identify it--PCR starts with a "primer", or target sequence. If the target is present in the DNA, a chain reaction of replication begins until the target sequence reaches measureable levels.
Shorter and less technical version: They look for a short stretch of DNA known to correspond to that of a GMO. In other words, they look for those specific DNA sequences that were inserted into the genome of the GMO by the biotech company that made them. They aren't trying to detect "design" in some abstract sense, they're looking for a specific sequence with a known origin.

There's a lesson here. In the real world, design detection does not work by the Dembskian method. We don't decide that something is designed simply by eliminating any and all competing hypotheses without any regard for how likely or unlikely the design hypothesis is itself. What we do instead is compare hypotheses, and only when the design hypothesis is itself sufficiently supportable are we willing to consider it a contender. And before we can do this, we must have a specific hypothesis that can be evaluated on its own merits, not merely a vague attribution of "design". In our present case, there are specific sequences that are known ahead of time to have been designed by human beings. So if we detect such sequences, then our particular design hypothesis (not just any design hypothesis) can be considered strongly supported. When weighed against competing hypotheses, such as the sequences appearing by chance in a genome where they're not normally found, the design hypothesis is clearly superior. It would be otherwise if we had no design hypotheses with any support.

That is a good lesson in and of itself, but there is deeper and somewhat amusing problem with Cordova's argument, one which makes it inoperative even if Genetic-ID did apply an eliminative method to detecting GMOs. Dembski and the entire ID movement argue that when the EF is applied to biology, it shows us that all the living things we find out in nature are "designed". And moreover, the EF (and by extension, ID theory as a whole) cannot detect the source of this design, it can merely tell us that living things are definitely designed by some agency or another. That being the case, how could this company, Genetic-ID, possibly be applying the EF? It would stymie them every time. According to Dembski, there are no organisms that aren't designed, so applying the EF would simply tell us the same thing over and over again, regardless of what organisms it was applied to. And we wouldn't know if the design was caused by human beings or by God, because the EF is inherently incapable of making that distinction. So if Genetic-ID took Dembski's ideas seriously, their job would be impossible. Far from being an example of how Dembski's EF is used in the real-world, this shows us why the EF cannot possibly have any real-world applications in biology. The only way it could be otherwise is if we accept that living things (with the exception of GMOs) are not designed. And that would sort of defeat the whole purpose of the ID movement.