Evolution Icon Evolution
Intelligent Design Icon Intelligent Design

The Evolutionary Informatics Lab: Putting Intelligent Design Predictions to the Test

Casey Luskin

One of the brightest spots of the intelligent-design research program, highlighted in our newly updated listing of pro-ID peer-reviewed publications, is the Evolutionary Informatics Lab. The lab’s founders, William Dembski and Robert Marks, have some of the strongest credentials in the ID movement. With PhDs in both mathematics and philosophy, Dembski is a leading light of ID. Marks is Distinguished Professor of Electrical and Computer Engineering at Baylor University and has over 250 scientific publications to his name, including many in the field of evolutionary computing.

The lab got off to a rough start in 2007 when Baylor administrators learned that Marks was doing ID-friendly research on the campus. A Baylor dean e-mailed Marks with the order that he “disconnect this [lab’s] web site immediately.” Before the thought police were done, Baylor forced the Evolutionary Informatics Lab not just to remove its website from university servers, but also to return a five figure grant. Universities aren’t known for turning down free money — but apparently the censors at Baylor preferred not have $30,000 for research if it might go to support intelligent design.

Despite the setbacks, the lab has attracted graduate student researchers and to date has published six peer-reviewed articles in mainstream science and engineering journals. In their papers, Dembski and Marks have developed a system for studying evolutionary algorithms — computer programs of digital organisms that, according to ID-critics, show that Darwinian processes can create new information.

Using their methodology Dembski and Marks have quantitatively measured the amount of “active information” smuggled into the evolutionary simulation by the programmer to allow it to achieve its goal. The analyses support “no free lunch” theorems — the notion that without intelligent input there can be no gain in complex and specified information.

Thus far, Dembski, Marks and their graduate research assistants have identified sources of active information in key evolutionary algorithms — “Avida” and “ev” — programs that have been widely touted as refuting ID. Their work shows that these programs do not truly model blind and unguided Darwinian processes, but cheat because they were pre-programmed by their designers to reach their digital evolutionary goals. As the lab’s website suggests, “Evolutionary informatics … points to the need for an ultimate information source qua intelligent designer.”

We’ve previously discussed many of the papers from the Evolutionary Informatics Lab here at Evolution News & Views (for example, see here, here, here, here, here, or here), but it’s worth highlighting some of the articles that have been newly added to our pro-ID peer-reviewed articles page.

Building the Methodology
One of the first papers published by the Evolutionary Informatics Lab came out in 2009, by William Dembski and Robert Marks in Proceedings of the 2009 IEEE International Conference on Systems, Man, and Cybernetics. Titled “Bernoulli’s Principle of Insufficient Reason and Conservation of Information in Computer Search,” the paper contends that in all searches — including Darwinian ones — information is conserved such that “on average no search outperforms any other.” The implication of their principle of “Conservation of Information” (COI) is that Darwinian evolution, at base, is actually no better than a random search.

To make their argument, the paper develops a methodology for measuring the information smuggled into a search algorithm by intelligence. Exogenous Information (I) represents the difficulty of a search in finding its target with no prior information about its location. Active Information (I+) is the amount of information smuggled in by intelligence to aid the search algorithm in finding its target. Endogenous Information (Is) then measures the difficulty the search will have in finding its target after the addition of Active Information. Thus, I+ = I – Is.

Having laid this theoretical groundwork, Dembski and Marks aim to apply their ideas to evolutionary algorithms that claim to produce new biological information. They argue that computer simulations often do not properly model truly unguided Darwinian evolution: “COI has led to the formulation of active information as a measure that needs to be introduced to render an evolutionary search successful. Like an athlete on steroids, many such programs are doctored, intentionally or not, to succeed,” and thus “COI puts to rest the inflated claims for the information generating power of evolutionary simulations such as Avida and ev.” They conclude that when trying to generate new complex and specified information, “in biology, as in computing, there is no free lunch,” and therefore some assistance from intelligence is required to aid Darwinian evolution find unlikely targets in search space.

Another paper from the Evolutionary Informatics Lab appeared in Journal of Advanced Computational Intelligence and Intelligent Informatics in 2010. Co-authored by Dembski and Marks, the paper argues that without information about a target, anything more ambitious than a trivial search is bound to fail: “Needle-in-the-haystack problems look for small targets in large spaces. In such cases, blind search stands no hope of success.” They cite “No Free Lunch theorems,” according to which “any search technique will work, on average, as well as a blind search.” However, in such a case, “Success requires an assisted search. But whence the assistance required for a search to be successful?”

Dembski and Marks thus argue that “successful searches do not emerge spontaneously but need themselves to be discovered via a search.” However, without information about the target, the search for a search itself is still no better than a blind search: “We prove two results: (1) The Horizontal No Free Lunch Theorem, which shows that average relative performance of searches never exceeds unassisted or blind searches, and (2) The Vertical No Free Lunch Theorem, which shows that the difficulty of searching for a successful search increases exponentially with respect to the minimum allowable active information being sought.” The implication, again, is that without the ultimate input from an intelligent agent — active information — such searches will fail.

Applying the Methdology
Having developed a methodology for measuring active information in computer simulations of evolution, the Evolutionary Informatics Lab then sought to apply these methods to various noteworthy programs.

In 2010, the lab published another paper titled “Efficient Per Query Information Extraction from a Hamming Oracle,” published in 42nd South Eastern Symposium on System Theory, which scrutinizes Richard Dawkins’s famous “METHINKSITISLIKEAWEASEL” evolutionary algorithm. They showed that it starts with large amounts of active information — that is, information intelligently inserted by the programmer to aid the search. The authors find that this form of a search is very efficient at finding its target — but only because it is preprogrammed with large amounts of active information needed to quickly find the target. This preprogrammed active information makes it far removed from a true Darwinian evolutionary search algorithm. The authors developed an online toolkit of programs called “Weasel Ware” for assessing these kinds of questions.

Another paper in which the Evolutionary Informatics Lab collaborators reported active information sources in a well-known evolutionary algorithm was authored by Winston Ewert, William A. Dembski, and Robert J. Marks II in Proceedings of the 2009 IEEE International Conference on Systems, Man, and Cybernetics. This paper scrutinized Avida, an evolutionary simulation program which had been published in Nature in 2003, and was touted by its creators as a refutation of ID which showed “that complex adaptive traits do emerge via standard Darwinian mechanisms.” But the team at the Evolutionary Informatics Lab refuted those claims. Their paper argues that Avida smuggles in “active information” to allow the simulation to find its evolutionary targets, such as the following:

  • “Active information from Avida’s initialization” where “[t]he initialization in Avida recognizes the essential role of the nop-C instruction in finding the EQU.”
  • “Mutation, fitness, and choosing the fittest of a number of mutated offspring.”
  • Most importantly, there is “Stair step active information” where the digital “mutations” in Avida are essentially pre-programmed to perform a useful function, and are rewarded for doing so.

Regarding this third point, the authors note that, “The importance of stair step active information is evident from the inability to generate a single EQU [the target function] in Avida without using it.” They ask, “What happens when no stair step active information is applied?” and note what the original authors of the Avida program wrote:

At the other extreme, 50 populations evolved in an environment where only EQU was rewarded, and no simpler function yielded energy. We expected that EQU would evolve much less often because selection would not preserve the simpler functions that provide foundations to build more complex features. Indeed, none of these populations evolved EQU, a highly significant difference from the fraction that did so in the reward-all environment.

But does real biology “reward” mutations to the extent that Avida does? The passage quoted above shows that when Avida is calibrated to model actual biology — where many changes may be necessary before there is any beneficial function to select for (irreducible complexity) — “none of these populations evolved” the target function.

Avida’s creators trumpet its success, but Ewert, Dembski, and Marks show that Avida uses “stair step active information” by rewarding forms of digital “mutations” that are pre-programmed to yield the desired outcome. It does not model true Darwinian evolution, which is blind to future outcomes and cannot use active information. The implications may be unsettling for proponents of neo-Darwinian theory: Not only is Darwinian evolution “on average… no better than blind search,” but Avida is rigged by its programmers to succeed, showing that intelligence is in fact necessary to generate complex biological features. Again, the lab’s investigators developed an online toolkit of programs called “Minivida” for dissecting this program.

Finally, also in 2010, Dembski and Marks, along with George Montañez and Winston Ewert, co-published a peer-reviewed paper in BIO-Complexity titled “A Vivisection of the ev Computer Organism: Identifying Sources of Active Information.” This study scrutinized Thomas Schneider’s “ev” program, well-known in the field of evolutionary computing, which has been widely cited as showing that Darwinian processes can increase information. But Evolutionary Informatics Lab showed that contrary to such claims, ev is in fact rigged to produce a particular outcome. According to the paper ev “exploit[s] one or more sources of knowledge to make the [evolutionary] search successful” and this knowledge “predisposes the search towards its target.” They explain how the program smuggles in active information:

The success of ev is largely due to active information introduced by the Hamming oracle and from the perceptron structure. It is not due to the evolutionary algorithm used to perform the search.

As the paper shows, active information is smuggled into the fitness function used by the ev program. Rather than showing that information can arise by Darwinian evolution, ev shows that intelligence is required. (Here also, the lab released an online toolkit of programs called “Ev Ware” for scrutinizing this program.)

The work of the Evolutionary Informatics Lab demonstrates that ID proponents are capable of producing innovative techniques for tackling questions related to intelligent design and evolution. First, the lab developed a methodology for studying the degree to which information is smuggled into evolutionary algorithms. Then, the researchers applied that methodology to various well-known programs like ev, Avida, and Dawkins’ “Weasel Simulation,” and successfully identified sources of “active information” in each. As the lab’s website promised, their research has shown that even the best efforts of ID-critics cannot escape the fact that intelligence is required to generate new information.


Casey Luskin

Associate Director, Center for Science and Culture
Casey Luskin is a geologist and an attorney with graduate degrees in science and law, giving him expertise in both the scientific and legal dimensions of the debate over evolution. He earned his PhD in Geology from the University of Johannesburg, and BS and MS degrees in Earth Sciences from the University of California, San Diego, where he studied evolution extensively at both the graduate and undergraduate levels. His law degree is from the University of San Diego, where he focused his studies on First Amendment law, education law, and environmental law.



Baylor UniversityDarwinEvolution SundayGod and Evolutionintelligent designpeer-reviewreligionRobert J. MarksWilliam Dembski