Critique of Dembski's
"Intelligent Design is Not Optimal Design"

By Wesley R. Elsberry

Posted May 6, 2002

William A. Dembski continues his series of essays advocating an Intelligent Design Creationist (IDC) stance. In this one, he recycles Paul Nelson's arguments concerning evolutionary arguments about optimality and suboptimality. Dembski considers three categories of design: apparent, intelligent (or actual), and optimal. Apparent design is the limited (according to Dembski) class of design which can be accomplished via natural processes unaided by agents. Intelligent design is accomplished by an agent, but not necessarily an infinitely intelligent agent. The term "actual" as applied by Dembski has a history of juxtaposition with "apparent" in his past essays, but in this essay the flavor seems to draw more from Dembski's description of intelligent designers making realizable designs in the presence of conflicting constraints. Optimal design as described by Dembski refers to a class of design which ignores the prevailing constraints that limit actual design, and thus can only be considered to apply to some ideal plane, not to any reality in which agents operate.

There are a variety of problems in this latest essay. As implied above, terminological issues pose not inconsiderable problems of their own. Part of the problem is that the topic of optimality is not simply a tabula rasa awaiting the impress of an IDC intellect to shape it. A quite considerable literature exists which explores various aspects of optimality as it applies to biological issues. As the literature on optimal foraging theory makes clear, optimality is to be considered as finding the best approach with respect to a set of relevant constraints. In other words, "optimality" as it is already used by biologists corresponds rather closely to the description that Dembski gives for "constrained optimization". "Constrained optimization" is touted by Dembski as a feature of designs produced by intelligent designers and thus characteristic of actual design. But now we have the curious situation where the theologian climbs the mountain - and finds the biologists already ensconced there. The "optimality" discussed by biologists is "constrained optimization", and thus shares the properties that Dembski asserts for "actual design".

The deployment of arguments that stray into theological areas is a matter of concern for Dembski. The mere fact that scientists do make arguments based upon theological themata is taken as a weakness by Nelson and Dembski. But the intrusion of theological issues into the scientific arena is not simply a matter of ancient history. To make arguments that plainly seek to advance theological goals in relation to science and then critique scientists for having the temerity to venture a response on theological turf seems to me to have the flavor of chutzpah. What is apparently galling to Nelson and Dembski is that the scientists have been so effective when making those theological arguments.

Dembski proceeds to bash Stephen Jay Gould for Gould's critiques which employ considerations of optimality, attributing to Gould the use of "optimality" as referencing some unachievable ideal design. This is a fundamental error, as Gould is not doing this. Gould's use of optimality is consistent with other biological usage, and thus comports with the real-world category of "constrained optimization". Gould, Dembski claims, cannot critique a designer without having knowledge of the objectives of the designer. This is erroneous. There exist valid arguments which make designers susceptible to criticism on the basis of their designs. Specifically, one can reliably infer inconsistency on the part of a designer whose range of designs incorporate modules whose functional efficiency may vary widely, but which are deployed to achieve the same functions in different organisms. Dembski offers an appeal to ignorance as a defense against this kind of argument. According to Dembski, we cannot know that a particular design could actually be improved, or that putting the improvement in place would not create deficiencies elsewhere. This is a particularly weak defense. Consider the case of the panda's thumb, which Dembski addresses in his essay. If the same designer produced both the thumb in primates and the thumb of pandas, vague hints that problems might arise were the panda to be fitted with a primate-style thumb seem ineffective as a rejoinder. If the same designer is responsible for both designs, that designer is inconsistent. Period.

Tom Scharle has developed this sort of argument in posts on talk.origins. Instead of concluding inconsistency on the part of a single designer, though, Scharle uses such a finding as support for the notion that multiple designers are responsible. What Intelligent Design Creation conjectures seem to naturally lead to is a kind of polytheism, or at least poly-designer-ism. The plethora of design approaches to solve the same functional problems in biology bespeaks a similar plethora of designers, or a designer with a serious case of multiple personality disorder. What it does not bespeak is a single designer utilizing the concept of "intelligent re-use".

Dembski claims that suboptimality arguments have gained their appeal by abandonment of a scientific approach and adoption of a theological approach. Dembski says that the question of how could an existing structure be improved is set aside in favor of asking what sort of God or designer would create such a structure. This is, I believe, a misrepresentation. The question of how a particular structure could be improved is almost always addressed, although it may be more implicit than explicit in some cases. In the case of discussion of thumbs, for example, we have a variety of different thumb designs that are actualized. We don't have to wonder *how* a panda's thumb might be improved; we know from the digits that operate the space bars of our typewriters and computers where an improvement can be found. That it is an improvement could only be denied by the ideologically committed. On measures of range of motion, flexibility, and strength there is no question of its superiority. As for the question of what sort of God or designer would create such a structure, I believe Dembski has not quite formulated the correct question. Instead, the question asked is what sort of designer, having designed both structure A and structure B having shared function X where structure A is clearly superior to structure B, decides to actually deploy structure B anyway. The answer is, plainly, an inconsistent designer.

Dembski again touts "reliable indicators of design". There is an almost desperate quality to the race to distance design detection from anything that might lead to consideration of attributes of the designer. But the fact is that the "reliability" is simply asserted rather than demonstrated. When no information is available concerning the actual cause of an event, Dembski's "complexity-specification" criterion is completely powerless to distinguish between that event being the product of natural causes (having the "appearance of design") and the event being the product of agent causation (having the attribute of "actual design"). (See my previous criticisms of Dembski's META essays, linked from The Anti-Evolutionists: William A. Dembski.)

Dembski's proposed connection between suboptimality arguments and the problem of evil ignores the actual power of suboptimality arguments. The mere existence of systems which display "constrained optimization" does not argue for intelligent agents at the expense of natural selection. Dembski identifies "choice" as the hallmark of intelligent agent causation, but ignores the fact that "selection" is a synonym for "choice" in this regard. Natural selection, it turns out, has all the properties that Dembski claims for intelligent agents in "The Design Inference". It should surprise no one that the Design Inference has no means of distinguishing events due to intelligent agency from those due to the operation of natural selection.