Monday, October 23, 2006

"But is it specified?"

Before I try to see whether the Universal Probability Bound has any relevance to biological systems, I would like to look at another issue. Given any sequence of cards in the example above – or any set of data derived through either design or a chance process, is it possible to establish that it definitely isn't specified? More generally, does the idea of specification that Dembski uses have any real meaning at all? The commenters on my earlier post suggest not.

Forget formal definitions for now, and notice first that intuitively, specification does have some meaning. If I look at a sequence of playing cards (as before), I will come to some conclusion as to whether it is a “specified” sequence. If the cards are set out and there is no obvious pattern, then I will just shrug at the magician who has set them out. If they are set out, and an order is evident, then I will assume that this is due to the cleverness of the magician. In both cases, I have some idea of what I am looking for in terms of specification, even without the magician speaking. For most magic tricks, the magician would specify exactly what card or cards we are expecting to see – perhaps a card we pulled out earlier, even if we don't know how the magician knows it. In this case, the specification is obvious. But even without this, if the cards are laid out in an obvious order having supposedly been shuffled beforehand, then we will be convinced that this order isn't a chance arrangement, but present due to the intention of the magician.

However, it is possible that a magician will have some subtle specification that he has to explain to me, as it isn't otherwise obvious. For example, the cards may be ordered so that if they are set out going round in a spiral, the pattern becomes visible. A part of the challenge to Dembski's proposals is that I can't exclude the possibility that data has some specification that I don't know about. Sequences of data can intuitively fall only into one of two classes:

1)Data with an obvious specification
2)Data without an obvious specification

Here's another example, to illustrate. Here is a sequence of letters:


They look random – they obviously don't say anything in English – but how do I establish that they are not specified? How do I know that they aren't saying something in English in a cipher? How do I know that they don't say anything in French in a cipher? In an obscure language spoken by only 500 people in the South Pacific? In a highly advanced code? It doesn't have an obvious specification, but I'm not able to conclude that it doesn't have a hidden specification.

What impact does this have on our consideration of the UPB?

Dembski argues that the significance of the UPB is that if a specified event is less probable than the UPB, then it is unreasonable to conclude that this is a chance occurrence. I think it is fair to say that, whilst people objecting to Dembski's proposals may not be happy that the concept of specification has been defined in watertight terms, it is quite clear that the cases with which he is concerned are nonetheless specified – and objections on the basis of what happens in unspecified events of low probability - like the hand obtained when dealing a shuffled pack of cards - are therefore not sound.

Consider a hypothetical new protein, novofunctionase (nfu), which enzymatically switches cis-reactol to trans-reactol. For the the function of nfu to be expressed, let's say that DNA must code for of a precise sequence of 50 amino acids. There is the specification, then. I haven't written a general definition of specification, but I have given a specification of what is required for this protein to appear. The specified event that would cause this new protein to appear when it was not present before is the random appearance of DNA that codes for this sequence of amino acids. The probability of such a sequence appearing in a random sequence of DNA would probably be of the order of 1 in (3/64)50 – that is, 3.5x10-67. Supposing instead we were talking about a precise sequence of 200 aa's – a much more demanding specification. The probability of such a sequence appearing in a random sequence of DNA would be well below the UPB – therefore the UPB says that the appearance of such a sequence (as a random event, with no precursor, and not from an intermediate) could not reasonably be classed as a “natural” event – it would require “design” or “intent”. That is the idea that Dembski is trying to get across. I think that Dembski's terminology is sufficiently precise for this to convey a meaningful proposition at least in particular cases, even if we can't define a general case.

In this case, I have defined the specification – I have said what is necessary for nfu to be expressed. My specification here is very tight - “a precise sequence of 50 amino acids”. However a more general specification would be possible, as the functionality of nfu might be expressed by a variety of proteins with completely different sequences. Such a specification might look as follows:

The protein would require a sensitive area of 6 amino acids from particular groups of amino acids, to be sensitive to the presence of cis-reactol in the cell, and cause it to bind with the active site on the enzyme. It would require another area including 10 specific amino acids that would utilise an ATP molecule to allow the enzyme to switch a binding cis-reactol to trans-reactol, and then unbind it. It would require another 6 specific amino acids as gene markers, and it would require 8 specific amino acids to allow this gene to be switched on at the appropriate time in the life cycle of the cell. This specification is more detailed, and yet less specific than “a precise sequence of 50 amino acids”. But it is still a specification.

For the appearance of new functionality, specification is required. The specification may be hardly anything at all – maybe a sequence of almost any three amino acids will have a catalytic effect on a range of biochemical reactions – but there is still a specification. The fact that it is not possible to tell that an event is an unspecified event of low probability, rather than a specified one, is not relevant, because all the biologically significant events that Dembski is interested in are specified events.

To come ... looking at the specification of some other events