But Is It Deception? "The Deniable Darwin" examined...

David Berlinski writes:

[Quote]

Evolution is a process, one stretching over four billion years. It has not been observed. The past has gone to where the past inevitably goes. The future has not arrived. The present reveals only the detritus of time and chance: the fossil record, and the comparative anatomy, physiology, and biochemistry of different organisms and creatures. Like every other scientific theory, the theory of evolution lies at the end of an inferential trail.

This sort of confusion does not bode well. While there are evolutionary processes, evolution is itself a class of phenomena with characteristic diagnostic attributes. Biologists typically forward the change in allele frequencies or distributions in a population as an objective and quantifiable criterion. Ernst Mayr urges a more accessible set of criteria, specifically changes in adaptation and diversity. Unfortunately, the gain in common understanding comes at the price of losing clearly objective and quantifiable attributes.

Evolution as biologists define it has been observed. That is, there exist instances of phenomena which meet the objective and quantifiable criteria presented. These range from reversible changes in allele frequency as seen in peppered moths up through observed speciation events (some of which are considered to have crossed higher taxa).

There is no one grand theory of evolution. Evolutionary mechanism theories are legion; some are picayune, while others are of grand scale. Berlinski appears to conflate the theory of common descent with the term "evolution". This is a common confusion among theistic anti-evolutionists.

The present provides more than just grist for an inferential mill supporting common descent. It provides us with examples of evolutionary phenomena underway in the readily observable here-and-now. We can detect natural selection at work; we can measure genetic drift and gene flow in populations; and we can see the event that Darwin strove to explain, but never himself witnessed, the production of two species where once there was but one. The theory of common descent does lie, as other scientific theories do, at the end of a trail of inference. However, other evolutionary mechanism theories require but small amounts of inference comparatively and come with copious supporting observations. 

[Quote]

The facts in favor of evolution are often held to be incontrovertible; prominent biologists shake their heads at the obduracy of those who would dispute them. Those facts, however, have been rather less forthcoming than evolutionary biologists might have hoped. If life progressed by an accumulation of small changes, as they say it has, the fossil record should reflect its flow, the dead stacked up in barely separated strata. But for well over 150 years, the dead have been remarkably diffident about confirming Darwin's theory. Their bones lie suspended in the sands of time-theromorphs and therapsids and things that must have gibbered and then squeaked; but there are gaps in the graveyard, places where there should be intermediate forms but where there is nothing whatsoever instead.(1)

[(1) A.S. Romer's Vertebrate Paleontology (University of Chicago Press, third edition, 1966) may be consulted with profit.]

This rather neatly ignores the fact that we have excellent reasons to believe that there are both geological and biological factors that would argue strongly in favor of the kind of evidence that we do find in the fossil record. In fact, Darwin laid the question before his readers and proceeded to answer the objection in his usual clear and meticulous way. It is unfortunate that few in these latter days seem to have the wit to read Darwin for content rather than for text-bytes, for Darwin's list of reasons why the fossil record is and will remain incomplete stands the close scrutiny of modern understanding quite well.

George Gaylord Simpson once remarked that the remarkable thing about transitional forms is that they are always lacking -- except where they have been found. While anti-evolutionists like to stress the incompleteness of the fossil record, the fact is that even with such an incomplete record as we are left with, the available evidence discomfits entirely the notion of "special creation" (fiat creation of each and every species). This leaves theistic anti-evolutionists with yet another set of shrinking gaps in which to house their deity. Where special creation would have had to account for 5 billion separate creations, modern creation conjectures aim for slightly more modest figures, ranging from a few separate creations to a few thousand. Standard evolutionary change is invoked to fill the difference. The modern intelligent design conjecture disclaims the need to argue over how many original species or "kinds" might have been created; it is sufficient to the ID supporter to postulate that God might have provided feature upgrades while the evolutionary program was running. 

[Quote]

Before the Cambrian era, a brief 600 million years ago, very little is inscribed in the fossil record; but then, signaled by what I imagine as a spectral puff of smoke and a deafening ta-da!, an astonishing number of novel biological structures come into creation, and they come into creation at once.

The Cambrian does provide a bit of a puzzle, but not quite the one that Berlinski represents above. I personally like my "at once"s to refer to events significantly shorter than ten million years. 

[Quote]

Thereafter, the major transitional sequences are incomplete. Important inferences begin auspiciously, but then trail off, the ancestral connection between Eusthenopteron and Ichthyostega, for example - the great hinge between the fish and the amphibia - turning on the interpretation of small grooves within Eusthenopteron's intercalary bones. Most species enter the evolutionary order fully formed and then depart unchanged. Where there should be evolution, there is stasis instead - the term is used by the paleontologists Stephen Jay Gould and Niles Eldredge in developing their theory of "punctuated equilibria" - with the fire alarms of change going off suddenly during a long night in which nothing happens.

I find the clause, "Where there should be evolution," particularly troublesome. By whose reckoning should evolution be found there?

Even Charles Darwin knew better than that, stating quite clearly that natural selection operated intermittently, often only at long intervals. Strawmen do not good arguments make. 

[Quote]

The fundamental core of Darwinian doctrine, the philosopher Daniel Dennett has buoyantly affirmed, "is no longer in dispute among scientists." Such is the party line, useful on those occasions when biologists must present a single face to their public. But it was to the dead that Darwin pointed for confirmation of his theory; the fact that paleontology does not entirely support his doctrine has been a secret of long standing among paleontologists. "The known fossil record," Steven Stanley observes, "fails to document a single example of phyletic evolution accomplishing a major morphologic transition and hence offers no evidence that the gradualistic model can be valid."

What Stanley is talking about is the failure of "phyletic gradualism". Unfortunately, once Eldredge and Gould defined "phyletic gradualism", they and others started dropping the modifier "phyletic" without notice, assuming that the reader could disambiguate from context. This led to many who wished to see gradualism (the non-saltational transformation of populations) disputed simply reading these passages as if that were the topic being critiqued. It isn't. Phyletic gradualism is a distinct and distinguishable concept. 

[Quote]

When I observed that Richard Dawkins was unable to write a computer program that simulated his linguistic thought experiment, I did not mean that the task at hand was difficult. It is impossible. Mr. Wadkins commends the discussion in Keen and Spain's Computer Simulation in Biology as a counterexample; it is no such thing. What Keen and Spain have done is transcribe Dawkins's blunder into the computer language Basic. Here are the steps they undertake. A target sentence is selected -- basic biological modeling is fun. The computer is given a randomly derived set of letters. The letters are scrambled. At each iteration, the computer (or the programmer) compares the randomly derived sequence with the target phrase. If the arrays -- sequences on the one hand, target phrase on the other -- do not match, the experiment continues; if they do, it stops.

There is nothing in this that is not also in Dawkins, the fog spreading from one book to the next. The experiment that Keen and Spain conduct is successful inasmuch as the computer reaches its target; but unsuccessful as a defense of Darwinian evolution. In looking to its target and comparing distances, the computer is appealing to information a biological system could not possess.

This point seems to be less straightforward than I imagined, so let me spell out the mistake. Starting from a random string, suppose the computer generates the sequence bndit disne sot sodiswn toswxmspw sso. Comparing the sequence with its target, it proposes to conserve the initial "B." But why? The string is gibberish. Plainly, the conservation of vagrant successes has been undertaken with the computer's eye fixed firmly on its future target, intermediates selected not for what they are (gibberish, after all), but for what they will be (an English sentence). This is a violation of the rule against deferred success. Without the rule, there is nothing remotely like Darwinian evolution. What the computer has in fact done is to match randomly selected items to a template, thus inevitably reintroducing the element of deliberate design that was banished from the Darwinian world.

If the task that Berlinski claims is impossible is the generation of sentences without a template by means of the principles of natural selection, then he is just dead wrong. One can eliminate the "distant ideal target" easily. What, then, is the information that can be used? Berlinski indicates that our program should have no information that is not available to a biological system accomplishing the same task. In the case of language production, the biological systems which do that are individuals who have the examples of language used in their presence by others. It is thus legitimate to give information concerning language usage to our computational system, just not a "distant ideal target" as a template.

One can find a suitable repository for such information in the n-grams of Shannon's "approximations to English". Instead of being given rules of grammar and syntax, we can provide an even lower level of information for our program, meaning that generated phrases are constructed without any explicit reference to words, grammar, syntax, and especially semantics. One doesn't need to go as far as a genetic algorithm in order to generate a bunch of text with a superficial similarity to the English encountered on the 'net; the November 1984 Byte magazine featured a "travesty" program. The Perl "camel book" comes with a "travesty" generator. The Emacs text editor features the "dissociated press" script. Each of these, given a text to analyze for its character frequency and association content, produces text that is often novel, commonly grammatical, and even sometimes at least partially coherent.

For example, "dissociated-press" comes up with the following when fed Berlinski's quote above:

What the computer program that they are (gibberish, after all), but for what they are the steps they undertake.

Plainly, the experiment that simulated not match, the experiment succes on the one hand, target phrase on the task at hand was banished from the initial "B." But why? The string is nothing in this the letters are scrambled. At each iteration. In lookins, the for what they are target, intermediates selected not for what they undertake.