Sunday, December 17, 2006

Specification all the way down

Many people have engaged with Richard Dawkins on the issue of METHINKSITISLIKEAWEASEL, the computer program he wrote that seemingly models an evolutionary process. Probably the most ... well, meaningful, is found in “A Meaningful World”, by Benjamin Wiker and Jonathan Witt. This unpacks the significance of the line that Dawkins chose in its Shakespearean context, to demonstrate that the almost complete lack of congruence between the way in which Shakespeare used language (and we respond to language, as demonstrated even by Dawkins' choice of phrase for this test) and a random walk towards a target phrase.

It has struck me that there are many levels of specification in language – this came up in an earlier discussion when we talked about the likelihood of a sequence of letters chosen at random conveying information. Perhaps the complexity of specification of language is the reason that Dembski didn't (as far as I know) use language as an example of specification in his books.
Let's set aside issues of the case of letters (which represents another level of specification) and punctuation (which represents yet another), and just consider the levels of specification involved in series of letters. What levels of specification can we list?

Level 0 – from all possible symbols chosen to write with, pick those that constitute the character set used in the English language.

Level 1 – letters should be organised in such a way as to represent a text – aligned, and flowing in a regular direction.

Level 2 – words must be correctly encoded:
Disallows DEF YUOH GBY
Allows DEN YOUR GUY

Level 3 – words must be organised in grammatically reasonable sequences:
Disallows CABIN THIS RED OPENS JUMPER
Allows THIS RED CABIN OPENS A JUMPER

Level 4 – collections of words must convey meaning:
Disallows THIS RED CABIN OPENS A JUMPER
Allows THIS RED KEY OPENS A DOOR

Level 5 – collections of collections of words must convey meaning:
This level encapsulates what we would consider to be logic – two “sentences” which contradict one another should not be present in the same text unless something happens between the two sentences to allow this to make sense. Thus, “God is infinitely mutable in his essence. God is infinitely immutable in his essence,” is only a reasonable text if you throw away the concept of logic. It sounds profound, but in actual fact, it is undermining the common logical currency that we share. It only makes sense if language doesn't make sense.

Note that works of fiction will generally take place at this level – that is, a work of fiction ought to be well-specified at this level – it should not contain internal contradictions.

At a high level, I guess, is the idea that there ought to be a correspondence between the text and the reality denoted. So the truth of THIS RED KEY OPENS A DOOR now hangs on a relationship that has to be established between the phrase and a specific red key – one that actually exists. A text at this level ought to be consistent with the worldview of the author.

At the highest level is the idea of TRUTH. I don't see anything unreasonable with the idea that a text can be congruent with reality at a level that exceeds our ability as observers to apprehend it. This is my understanding of the nature of the Bible. It would have to be written from a perspective that was higher level than any human's - again, this is my understanding of the nature of the Bible.

This ties in with the “Data/Information/Knowledge/Wisdom” idea expressed in the headline above. Many of these levels of specification also have analogues in the realm of proteins and DNA. A gene defines a function (ultimately) – that is, a gene provides an organism with the information it needs to solve a particular problem. There is a defined alphabet of four nucleic acids, and they have to be grouped in threes to define an amino acid. But a protein can't contain any old sequence of amino acids – it can't even have the right amino acids in any order. In addition to the amino acid sequence (the primary structure), there is also the secondary structure, which includes things like alpha helices, and the tertiary structure, which is the overall shape of the protein unit, and even the quaternary structure, which is how multiple protein units might function together. All of these are different levels of specification, like the different levels of specification in language that allow a series of symbols on a page to convey meaningful information to a reader.

With the exception of onomatopoeic words, there is generally little connection between the words that we use and the concepts that those words contain. The abstraction of concepts from the universe to symbols in languages – and the ability to move from this abstraction to create entirely new ideas – is one of the defining characteristics of humans. A dog may learn what “walk” means, and non-human primates and dolphins may learn the names of objects – but this shift from concrete to abstract, and then the realisation of abstract in concrete again are uniquely human.

In these multiple layers of specification, there are also echoes of things such as the Open Systems Interconnection seven layer model and the nature of mathematics, where philosophers are quietly amazed that there should be a connection from the abstract world of numbers to the real world of countable objects.