The phrase "complex specified information" is a core concept in the lexicon of the World's Greatest Information Theorist William Dembski, but it was not invented by him. In fact, it comes from a book published in 1973, by Leslie Orgel, entitled The Origins of Life: Molecules and Natural Selection. It is odd to think that this notion, which is supposed by Dembski to demonstrate that life, or LIFE as he like to refer to it in his The Design Inference of 1998, cannot be the result of natural selection, was coined in the context of a discussion how it actually could.
What does Orgel actually say about complex specified information, then? He didn't actually use the complete phrase "complex specified information", but the ideas are there. Perhaps in reading him, we may get a hint of how Dembski developed his reaction to Darwinian evolution, and more to the point, perhaps we'll be able to make sense of some wider, and more important, ideas in evolution.
Orgel's book is an introductory discussion to the state of Origins of Life research and ideas at the date of writing. He first describes the history of debates, including the spontaneous generation debate, the development of biochemistry and theories about the formation of the solar system, and the ways in which natural selection could act at the molecular level. He notes that life presents an apparent paradox of how things could show design and purpose, and how Darwin's theory of evolution by natural selection undercut this (p181):
Darwin's work ... showed how complex well-adapted organisms evolve from organisms that are simpler and less well-adapted. Recent experimental studies fully justify his belief that the variations that make selection possible are the consequences of random events. Why do so many philosophers claim there is a paradox about evolution?
The supposed paradox is concerned with design. ...
Orgel discusses the need we have to use "loaded" words that suggest intention, but that these are abbreviations of non-loaded terms, a point noted by Dawkins later on. Yes, natural selection is counterintuitive, but intuition is a poor guide - we do not do science on the basis of how we feel.
After this Orgel discusses whether or not we will find life on other planets - something that is rather pointed as we watch the results come in from the Mars rovers. To answer this, he has to define what he means by "life" in ways that do not depend on the contingent properties of life on earth. Of course, this means he gets a little... abstract. What he defines life as is CITROENS. Not the car, but a class of things that do things. What things? This is where complexity enters the building.
We can try to define life from nonlife in terms of chemistry or behavior, but the best way is in terms of molecular structure and behavior (p189):
In brief, living organisms are distinguished by their specified complexity.
In a footnote, Orgel says
It is impossible to find a simple catch phrase to capture this complex idea. "Specified and, therefore, repetitive complexity" gets a little closer..."
To illustrate this, he compares life to crystals (something Pasteur also did) - crystals fail to qualify as living, he says, because they lack complexity; mixtures of polymers because they lack specificity. Here's where Orgel goes into ambiguous territory, though. To explain complexity and specification, he makes use of the notion of "information" (p 190). Structures are specified by "instructions". Living things need to be specified by complex instructions.
A moment's thought would indicate that Orgel is using a metaphor. Crystals are not specified by instructions, nor are polymeric configurations. They are caused by the physical properties of molecules packing according to the laws of thermodynamics, to achieve the least energetic state. But the information metaphor is pervasive, and is, I think, another example of the same mistake Orgel was discussing above, and is a loaded word. It was very popular among scientists working with the difficult field of biology at the time. Indeed, the so-called Central Dogma of genetics is that "information" is never transferred from proteins to DNA. But all that means is that the arrow of cause and effect runs from DNA to proteins - DNA molecules cause proteins to be formed in a sequence, but proteins do not cause DNA to be formed in a sequence.
But the metaphor is taken literally by many. Despite qualifying what he says of genes and replicators, Dawkins is taken in by this loaded word too. And what confusion it is causing, too, nowhere more than in William Dembski's asseverations.
Orgel is careful not to fall into this trap too far. His discussion of why a solution of polymers has complexity but no specificity resolves to - polymer sequence is random, DNA in two identical E. coli is identical and therefore not random. Specificity is, in effect, the inverse of randomness.
Of course, the conformation, as it is called, of any molecule is anything but random. The laws of chemistry are highly specific, and you get the same molecules forming in the same conditions - that's what a chemical reaction is. But let's move on to Orgel's CITROENS.
We agree, he says, to call anything alive that exhibits some fundamental properties: reproduction, metabolism, excitability, and so on (p192). He notes that reproduction involves the structures of the children being the same as those of the parents - this is "information" being "passed on". And so he sets up the following claim (p193):
These considerations lead us to the following requirements that are necessary and sufficient to qualify a structure as "alive":1. The object is complex and yet well-specified.
2. The object is able to reproduce."
These conditions ... imply that:
(a) the object is the product of natural selection,A new term for such "living" organisms, whether terrestrial or not, must now be introduced. They are Complex Information-Transforming Reproducing Objects that Evolve by Natural Selection - CITROENS.
(b) the information needed to specify the object is stored in a structure that is stable for the reproductive lifetime of the object,
Orgel ends by discussing - suprise, surprise - William Paley, and arguing that "the operation of natural selection, a completely random process, leads to the evolution of organisms that do seem to have been designed" (p196). Perhaps Dembski read Orgel's book when an undergraduate, and this stuck in his craw, or perhaps he thinks that the world ought to be as it seems. Who knows. But one thing is clear - CSI is nothing like the way he describes it.
Dembski's concept is pretty much arbitrary - you have to exceed a certain threshold of improbability for something to be "complex-sensu-Dembski". Moreover, "specified-sensu-Dembski" is not - as it is in Orgel's sense - a matter of causal transmission (which could be defined as "information, at least in the sense used in Communications Theory), but some vague notion of meaningful structure. Finally, Dembski is not talking about information in the usual ways, and certainly not in terms of heredity as Orgel was. Of the two, Orgel is far less confused and arbitrary and ambiguous. Richard Wein has discussed this in some depth at Talk.Reason. See also here, and here, and here.
Tom Schneider dissects Dembski's CSI here.
Wes Elsberry and Jeff Shallit discuss what Dembski's definition of "information" is and what it means here. [PDF document]