Science & Faith
9/7/13 at 10:21 PM 21 Comments

Evolutionary Computer Games, Natural Selection, and Intelligent Selection

text size A A A
Wiki Commons
Avida: Evolutionary Computer Simulation

A July 16, 2013 paper in the prestigious Proceedings of the National Academy of Sciences by Richard Lenski and colleagues, "Experiments on the role of deleterious mutations as stepping stones in adaptive evolution," seeks to establish a more important evolutionary role for harmful mutations by means of a computer simulation. Thanks goes to the National Science Foundation (your tax dollars) for funding Lenski's decade-long computer game (using "Avida" pictured here). What intellectual return have we received from this investment? Unfortunatetly, not much.

Here are some highlights of an interesting critique of Lenski's work over at ENV.
It's unsurprising to learn that Lenski & Co. have resorted to a favorite magic kit, the evolutionary Avida program, written by philosopher Robert T. Pennock, author of the anti-ID book Tower of Babel. We've discussed this and similar algorithms many times.

Whether the "digital organisms" generated by such programs have any connection to the real world is highly unlikely, especially when human beings are rewarding them by design. The digital organisms live as long as the programmers let them. They don't have to find food. They don't have to endure the slings and arrows of outrageous fortune. Anyway, what does a logic routine in a computer have to do with building a trilobite?

Computer programmers are intelligent designers. Although in this case they attempt to simulate unintelligent evolutionary change (mutation and natural selection unguided by intelligence), a significant degree of intelligent investigator interference still occurs.

In a similar mode of critical review ENV continues:

In Avida, the fitness of digital organisms is measured in terms of logic functions. Life, though, revolves around genes and proteins. There's no comparison. For one thing, the information content of the typical gene or protein is vastly greater than the information content of an "AND" or "NAND" or "EQU" logic gate. Proteins typically include hundreds of precisely sequenced amino acids, coded for by equal numbers of genetic alphabet letters (150 amino acids is a relatively small protein).

Moreover, a polypeptide sequence has to fold into a functional shape. A precisely ordered sequence is not enough. Without a stable fold, the polypeptide is useless, if not harmful. Just a few mutations to a working protein are often enough to destabilize the fold. An unstable protein will be targeted by proteases and destroyed; those that are not will be eliminated by "purifying selection." This has the effect of making the walls of a fitness peak steeper, and the valley deeper -- more like a cliff on the edge of an ocean filled with sharks.

Indeed, there are very few ways to be alive, but many ways to be dead. We can picture a vast ocean to represent possible amino acid sequences that don't amount to life-helpful proteins. In that huge ocean of non-functional sequence space are widely separated islands of function: amino-acid chains that fold and behave in a life-helpful manner. In addition, as the ENV article explains, those tiny islands have steep contours. This all indicates how the evolution of new biologically functional proteins would not happen (or almost never happen) given the probabalistic resources of our universe and the assumption that there was no intelligent guidance in those origin events.

Problem: Neo-Darwinism requires one to believe in an enormous number of such useful protein origin events. This is no problem in the intelligently designed computer world of evolutionary video games. But before Darwinists celebrate, they should recognize that the marvelous computer world selection of digital creatures has a critical component of intelligent selection (though perhaps subtly embeded in the backbround), and that it is not purely natural selection in the Darwinian sense.

If you wish to think more about natural selection itself, consider this other ENV essay: Natural Selection Is Empty. It opens with this:


Jerry Fodor and Massimo Piatelli-Palmarini's book What Darwin Got Wrong was published in 2010. Having read it now, perhaps belatedly, I can report that it is a masterpiece. Fodor is a leading philosopher, and Piatelli-Palmarini is a leading cognitive scientist. Their analysis of natural selection is meticulous and devastating. They are both atheists -- they do not come to this debate with theistic presumptions. They demonstrate that natural selection is, in their word, empty. It's a meaningless concept that should be abandoned.

F&P-P begin their argument with the observation that phenotypic traits on which natural selection acts are often linked at the genetic level. It is unusual to find a specific trait that can be selected without selecting for other traits. Cellular genetics is a complex interconnected affair. A change at the level of the gene generally has complex effects on phenotype.

When we say that natural selection acts, how do we know which phenotype is the object of selection, and which are free-riders? Preservation of one trait preserves linked traits. Gould and Lewontin recognized this dilemma. In their paper"The Spandrels of San Marco and the Panglossian Paradigm: A Critique of the Adaptationist Programme" they take to task the adaptationist (strict natural selection-ist) view that natural selection can select for specific phenotypes that are linked to other phenotypes. They call the ostensibly selected phenotypes arches, and the free-rider unselected phenotypes spandrels, referring to the difference between the structural arches in cathedrals and the decorated spandrels -- the spaces between the arches -- that serve an artistic, but no structural, purpose.

Gould and Lewontin point out, correctly, that phenotypes in nature are composed of arches and spandrels -- traits that enhance survival, and traits that are linked to survival traits genetically, but which provide no survival advantage themselves. They argue that adaptationist (natural selection-ist) explanations fail to take into account the fact that natural selection cannot distinguish between arches and spandrels, and that therefore invocation of natural selection, which is blind to the arch/spandrel dichotomy, is often an inadequate evolutionary explanation.

F&P-P unpack Gould's and Lewontin's critique, applying logical rigor. They point out that genetically linked traits are coextensive. Ya' select for one, ya' select for the other. They come as a package. F&P-P ask:

How can natural selection distinguish between, on the one hand, phenotypic traits that affect fitness and, on the other hand, their endogenously linked phenotypic correlates... selection [cannot] apply differentially to coextensive properties.

So merely invoking "natural selection" fails to provide an explanation for the survival of a trait, because natural selection is blind to the difference between traits that enhance survival and traits that are free-riders and irrelevant to survival, as long as the traits are linked. Read more.

Even if this empirical and conceptual problem is resolved in neo-Darwinism, countless other fundamental difficulties (see previous posts) render the theory implausible.

CP Blogs do not necessarily reflect the views of The Christian Post. Opinions expressed are solely those of the author(s).