The Evolutionary Informatics Lab, headed by Drs. William Dembski and Robert J. Marks, II, has published a number of peer-reviewed scientific papers that assess supposed simulations of evolution. Their team developed a methodology for studying so-called "genetic algorithms" -- computer programs that are intended to simulate the Darwinian process. These programs incorporate "active information," which is essentially the amount of information smuggled into a search algorithm by an intelligent programmer to help it find a target. Their methodology calculates the amount of active information in a program, showing that intelligence -- not Darwinian evolution -- is what is finding the targets of these searches. In a new peer-reviewed scientific paper in the journal BIO-Complexity, "Time and Information in Evolution," Winston Ewert, Ann Gauger, along with Dembski and Marks, once again show that a mathematical simulation of evolution doesn't model biologically realistic processes of Darwinian evolution at all. First, a bit about the paper that Ewert et al. are responding to.

The new paper responds to a 2010 paper in Proceedings of the U.S. National Academy of Sciences (PNAS) titled "There's plenty of time for evolution," by Herbert S. Wilf and Warren J. Ewens, a biologist and a mathematician at the University of Pennsylvania. There's little doubt that Wilf and Ewens intended their work to respond to the arguments of intelligent-design proponents. Though lacking any citations to ID literature, the paper's abstract starts off by stating, "Objections to Darwinian evolution are often based on the time required to carry out the necessary mutations." They then open the body of their paper by elaborating on these "objections":

The 2009 "Year of Darwin" has seen many welcome tributes to this great scientist, and reaffirmations of the validity of his theory of evolution by natural selection, though this validity is not universally accepted. One of the main objections that have been raised holds that there has not been enough time for all of the species complexity that we see to have evolved by random mutations. Our purpose here is to analyze this process, and our conclusion is that when one takes account of the role of natural selection in a reasonable way, there has been ample time for the evolution that we observe to have taken place.

But who exactly is making these "objections" that there isn't enough time "to carry out the necessary mutations"? People like Michael Behe, William Dembski, Douglas Axe, and Stephen Meyer have certainly made such arguments in peer-reviewed scientific papers published prior to the publication of Wilf and Ewens' paper. Conspicuously, however, Wilf and Ewens's paper cites none of those papers, nor any other literature making similar objections. This is a classic politically motivated tactic of Darwin defenders. You write a scientific paper attempting to respond to specific ID arguments, but then refuse to cite ID literature for fear of validating that there are legitimate scientific controversies at stake. But if there's "no controversy" over Darwinism, why is there a need to refute the arguments of critics in a prestigious journal like PNAS? Paul Nelson calls this "debating the controversy that doesn't exist."

In any case, the motivation behind Wilf and Ewens's paper wasn't lost on Darwin-bloggers. PZ Myers claimed the paper provided "a guide to short-circuiting the invalid assumptions of creationists." Jerry Coyne said it provides "one step towards dispelling the idea that Darwinian evolution works too slowly to account for the diversity of life on Earth today." According to Coyne, "what impresses me is the huge shortening of time that occurs under realistic assumptions."

But did the paper use "realistic assumptions"? According to the BIO-Complexity paper by Ewert, Dembski, Gauger, and Marks, the answer is a resounding "no." They offer a long list of reasons why the Wilf and Ewens model of evolution isn't biologically realistic:

Wilf and Ewens argue in a recent paper that there is plenty of time for evolution to occur. They base this claim on a mathematical model in which beneficial mutations accumulate simultaneously and independently, thus allowing changes that require a large number of mutations to evolve over comparatively short time periods. Because changes evolve independently and in parallel rather than sequentially, their model scales logarithmically rather than exponentially. This approach does not accurately reflect biological evolution, however, for two main reasons. First, within their model are implicit information sources, including the equivalent of a highly informed oracle that prophesies when a mutation is "correct," thus accelerating the search by the evolutionary process. Natural selection, in contrast, does not have access to information about future benefits of a particular mutation, or where in the global fitness landscape a particular mutation is relative to a particular target. It can only assess mutations based on their current effect on fitness in the local fitness landscape. Thus the presence of this oracle makes their model radically different from a real biological search through fitness space. Wilf and Ewens also make unrealistic biological assumptions that, in effect, simplify the search. They assume no epistasis between beneficial mutations, no linkage between loci, and an unrealistic population size and base mutation rate, thus increasing the pool of beneficial mutations to be searched. They neglect the effects of genetic drift on the probability of fixation and the negative effects of simultaneously accumulating deleterious mutations. Finally, in their model they represent each genetic locus as a single letter. By doing so, they ignore the enormous sequence complexity of actual genetic loci (typically hundreds or thousands of nucleotides long), and vastly oversimplify the search for functional variants. In similar fashion, they assume that each evolutionary "advance" requires a change to just one locus, despite the clear evidence that most biological functions are the product of multiple gene products working together. Ignoring these biological realities infuses considerable active information into their model and eases the model's evolutionary process.

(Winston Ewert, William A. Dembski, Ann K. Gauger, Robert J. Marks II, "Time and Information in Evolution," BIO-Complexity, Volume 2012 (4).)

"Because of these problems," the authors argue, Wilf and Ewens's "conclusion that there's plenty of time for evolution is unwarranted."

But the abstract hints at a fundamental problem with Wilf and Ewens's paper -- a problem that shows why it's not safe to assume as Wilf and Ewens do that mutations can always smoothly accumulate in a parallel fashion.

In Wilf and Ewens's evolutionary scheme there is a smooth fitness function. Under this view, there is no epistasis, where one mutation can effectively interact with another to affect (whether positively or negatively) fitness. As a result, any mutations that move the search toward its "target" are assumed to provide an immediate and irrevocable advantage, and are thus highly likely to become fixed. Ewert et al. compare the model to playing Wheel of Fortune:

The evolutionary model that Wilf and Ewens have chosen is similar to the problem of guessing letters in a word or phrase, as on the television game show Wheel of Fortune. They specify a phrase 20,000 letters long, with each letter in the phrase corresponding to a gene locus that can be transformed from its initial "primitive" state to a more advanced state. Finding the correct letter for a particular position in the target phrase roughly corresponds to finding a beneficial mutation in the corresponding gene. During each round of mutation all positions in the phrase are subject to mutation, and the results are selected based on whether the individual positions match the final target phrase. Those that match are preserved for the next round. ... After each round, all "advanced" alleles in the population are treated as fixed, and therefore preserved in the next round. Evolution to the fully "advanced" state is complete when all 20,000 positions match the target phrase.

Thus, Wilf and Ewens ignore the problem of non-functional intermediates. They assume that all intermediate stages will be functional, or lead to some functional advantage. But is this how all fitness functions look? Not necessarily. It's well known that in many instances, no benefit is derived until multiple mutations are present all at once. In such a case, there's no evolutionary advantage until multiple mutations are present. The "correct" mutations might occur in parallel, but the odds of this happening are extremely low. Ewert et al. illustrate this problem in the model by using the example of the difficulty of one phrase evolving into another:

Suppose it would be beneficial for the phrase

"all_the_world_is_a_stage___"

to evolve into the phrase

"methinks_it_is_like_a_weasel."

What phrase do we get if we simply alternate letters from the two phrases?

"mlt_ihk__otli__siaesaaw_a_e_."

Under the assumptions in the Wilf and Ewens model, the "fitness" of this nonsense phrase ought to be exactly half-way between the fitnesses of "all the world is a stage" and "methinks it is like a weasel." Such a result only makes sense if we are measuring the fitness of the current phrase by its proximity to the target phrase.

But the gibberish of the intermediate phrase doesn't cause any problem under Wilf and Ewens's model. Not unlike Richard Dawkins, they assume that intermediate stages will always yield some functional advantage. And as more and more characters in the phrase match the target, it becomes more and more fit. This yields a nice, smooth fitness function -- rich in active information -- not truly a blind search. Ewert et al. explain why this isn't anything like Darwinian evolution:

This example reveals two biologically unrealistic things about Wilf and Ewens's model. First, evolutionary processes can only depend on the performance of current organisms, not hypothetical target organisms. Only teleological processes have the ability to consider future phrases. Yet Wilf and Ewens's oracle has to have knowledge of the target to assess fitness. Second, the natural language example also highlights the importance of context for assessing fitness, if we think of fitness as the ability to convey meaning in a natural language.

Thus, just as Wilf and Ewens are wrong to assume that every intermediate phrase has meaning, so they are wrong to assume that every intermediate biological stage along some evolutionary pathway will be functional:

In addition to the overwhelming problems mentioned above, the search algorithm they have chosen is unrealistic. Wilf and Ewens assume that the fitness landscape is smooth, with each beneficial mutation trending upward additively. This is not the case in biology. ...Indeed, there is much evidence to suggest that real fitness landscapes have many local fitness optima surrounded by fitness deserts. If it takes more than several mutations to move from one peak to another, adaptation can become stalled on a local peak, with no way to move from one small fitness peak to a higher one. Because natural selection is blind and without foresight, it cannot tell which particular mutations are leading to an unrealized goal of maximal fitness (in this case a target phrase) some distance away in the adaptive landscape. It can only assess the relative local fitness of variants in the population.

In other words, Wilf and Ewens endowed their mathematical model of evolution with foresight. It is directed toward a target -- an advantage that natural selection conspicuously lacks. And what, in our experience, is the only known cause that is goal-directed and has foresight? It's intelligence. This means that once again, the Evolutionary Informatics Lab has shown that simulations of evolution seem to work only because they've been intelligently designed.

Image credit: Patrick Emerson/Flickr.