Although I have long been a critic of certain of William A. Dembski's claim, I find it interesting that in several respects I agree with Dembski in his latest META essay. I don't really care to quibble over whether the mechanics of recognizing the property or attribute of complexity-specification (which I shall refer to hereafter by its synonym, CSI) are proper to the last statistical detail. I will stipulate Dembski has provided a sufficient definition that can be used for identification of CSI in our discussions. Where we continue to part company is over the significance of CSI. Dembski's spirited defense of his inductive argument leading from identification of CSI as an attribute of an event to the implication of intelligent agency as a cause for that event is, nonetheless, not compelling in my estimation. I carried on an email exchange with Paul Nelson last summer over some of the points which are now being covered in Dembski's META essays. I will be using parts of my responses to Nelson in developing my further response to Dembski. There are some acronyms which I utilize which may not be apparent to the reader new to the discussion. I will list those here: DI: Design Inference (the argument itself) TDI: "The Design Inference" (Dembski's book about DI) CSI: Complex Specified Information NS: Natural Selection NTSE: "Naturalism, Theism, and the Scientific Enterprise" conference of 1997 Nelson had brought up a number of topics related to Design Inferences, and I made my responses in a topic by topic fashion. I will retain that structure here. Topic: Identification of proxies of intelligent agents by the DI. This takes us into issues of proximate versus ultimate causation. But I find my distinction to have merit in discussion of DI because such issues had been broached in TDI, as in the discussion of plagiary. If DI is insensitive to the difference between a CSI result produced by proximate action of an intelligent agent and one produced "by proxy" one or more steps removed from the proximate intervention of an intelligent agent, then interesting results fall out. The DI is thus incapable of distinguishing between the theological themata of Deism and any interventionist conception of God. The production of CSI by proxy does not tell us how many levels of indirection may lie between the proximate cause and the ultimate intelligent cause. This gets back to the vociferous "Imago Dei" argument between Foerst and Dembski at the NTSE. Interestingly, Foerst advocated a view in which the "Imago Dei" as an attribute could be passed on by humans to a robotic creation, and Dembski rather abruptly denied this view. But Dembski's defense of CSI as a marker of intelligent agency treats CSI as a particularly sticky genetic attribute: once it appears in a stream of information, the taint apparently never goes away. The whole issue of what it *means* to have a rational intelligent agent produce CSI is relevant here. I have objected forcefully to criticisms aimed at disbarring algorithms and natural process from being considered as sources of CSI when those same criticisms can (and should) be applied in equal measure to rational intelligent agents. Foremost in these considerations is how one measures the complexity of an event, object, or structure. Dembski's discussion of "detachability" in TDI yields the complexity measure of an event, object, or structure as the conditional probability of the event, object, or structure given some chance hypothesis H. In more recent essays, Dembski has attempted to replace this complexity measure when the causative agent is an algorithm. In that case, Dembski asserts that the relevant complexity measure is a conditional probability given a non-chance hypothesis. Dembski does not replace the chance-based complexity measure when an intelligent agent is believed to be the cause. This is simply insupportable. If one applied complexity measures based upon non-chance hypotheses, then nothing produced by an omnipotent omniscient entity could be considered complex, and very little produced by ordinarily intelligent humans could be so considered. Either the relevant complexity measure is the conditional probability given the chance hypothesis regardless of causation, or the entire DI becomes just another case of special pleading in an apologetic. Topic: Inability to contrast "evolutionary algorithms" with intelligent agents. I would ask, "Why not?" That a cause is complex opens the way to logical regress, but since the DI doesn't seem to care whether proximate or ultimate causation results in CSI, what does this matter? And then too, every intelligent agent we know anything about empirically is also complex, and yet this is not urged as a difficulty to consider any of them as a source of CSI. Taking this criticism and applying it to known intelligent agents yields yet another attack upon the concept of free will. If the CSI proximately caused by an intelligent agent can be attributed instead to the cause one or more steps removed of that intelligent agent, it would seem that the issue of who caused what CSI cannot even be approached using the DI, except perhaps within a Deistic theological framework. If we closely examine the position that an algorithm is a proxy or substitute for an intelligent agent, I think that we will find that places a pretty low bar for "intelligence". For example, a genetic algorithm can be analyzed to determine in just what fashion it "chooses between" actualized possibilities to exclude some and zero on on those which match a relevant specification. There simply is nothing in the GA which puts it beyond the "intelligence" that Charles Darwin identified as immanent in the processes of nature. Topic: No instances of CSI produced by natural selection are known. Let's explore CSI a little closer. We have two criteria that combine to yield Dembski's CSI: complexity at or above 500 bits and match to a specification. I will call this CSI_500 in order to introduce a slightly different perspective on CSI. Complexity is obviously distributed along the ordinal line of bits. Solutions to particular problems may represent CSI at a lower threshold than Dembski's step function of 500 bits proposes, but this should be clearly noted via use of a modifier to relate the level of complexity involved, like "CSI_250" or "CSI_50". Few question the ability of natural selection to produce solutions at lower complexity levels. (cf. Dembski's essay, "Intelligent design as a theory of information" from the NTSE.) Whether one admits "CSI_50", "CSI_250", or "CSI_499" as supported by the available evidence does not matter. What matters is that this level of performance is properly credited to the action of natural selection. Natural selection, though, is notoriously difficult to empirically isolate as a mechanism of action. The level of evidence needed to both implicate natural selection and to exclude genetic drift is high. Indirect evidence, such as the presence of linkage disequilibrium in a population, serves as an indicator of the action of natural selection, but biologists tend to want to see a clear relation between a cause of selection and an effect in distribution of traits in a population. Thus, the standard of evidence required for a *positive* and *exclusive* identification of the mechanism of change for many biological phenomena means that we cannot simply accept that some particular mechanism was at work for those cases where such evidence is unavailable. But this is different from an assertion that any particular mechanism was *incapable* of having produced the phenomenon in question. Taking it as possible that adaptive features of organisms are designed and installed by an intelligent agent via a mechanism other than natural selection means that we cannot use as examples of the efficacy of NS those phenomena in question, unless and until we have in hand the same kind of evidence that suffices for Galapagos finch beak changes. This may simply never be available. But if all that is available for the alternative hypothesis of ID is the simple fact of CSI_500, then I doubt that many biologists will feel compelled to exclude natural selection as a live possibility on those grounds alone. What we are then left with is an argument that we should exclude from consideration a mechanism of generating solutions that we can observe to happen in modern populations and which produces CSI at lower complexity levels during our brief and spotty periods of observation in favor of a mechanism which has no independent evidence of operation and which is not currently observable. (That is, the intelligent agent putatively responsible for the biological system under question is not known from current observation or from independent evidence of the period in question.) I think that such an argument will find it rough going to convince knowledgeable people of its merits. In order to explore this issue, it is necessary to calculate the CSI level of various examples of NS in action, or general "descent with modification" in action. Things like bacteria digesting nylon with novel enzymes or the emergence of the impedance-matching apparatus of the mammalian middle ear need to be explored quantitatively. This also applies to the putative examples of "design". While Dembski has offered Michael Behe's examples from "Darwin's Black Box" as instances of CSI ("Science And Design", First Things, Oct. 1998), there nevertheless has been no publication of the quantification that Dembski urges in TDI on p. 228: "Do the probability calculation!" A spread of CSI levels may indicate an approach to the CSI_500 level that Dembski sets, and indicate that no essential qualitative difference exists between the capability of natural evolutionary algorithms and intelligent agency. Without the quantification of specific instances, claims that DI is robust and reliable remain unsupported conjecture. The glib handling of probability by Dembski in deploying the "apparent CSI" / "actual CSI" split also indicates that the probability calculations that do get performed should be scrutinized carefully for preferential treatment of intelligent agents or prejudicial treatment of algorithms and natural processes. (See my critiques at , , and .) In other words, I find that CSI as given by Dembski is an attempt to turn a *quantitative* difference into a *qualitative* difference, and which I feel is itself a basic error. Step functions with arbitrary thresholds, even if the threshold has a post hoc rationalization attached, do not accomplish this task. If we are agreed that natural selection can increase information, then it would seem that there really is no bar to the production of CSI_500 via such means. If NS can be credited as the source of CSI_25 in one case and CSI_25 in another case, the incorporation of both into a single system could yield an instance of between CSI_32 and CSI_50 by reasonable information metrics (somewhere around CSI_32 being an expectation for a system composed of two conjoined copies of CSI_25 subsystems). Or NS working incrementally might advance the complexity of a system in a step-wise fashion, bit by bit. Given that NS can fix added complexity within a population, one would need to demonstrate a bar in principle to such complexity increases eventually exceeding the CSI_500 threshold set by Dembski. Topic: Robust character of CSI as a marker of intelligent causation. This should be a result of study, not an a priori position. The CSI_500 criterion apparently was chosen specifically to exclude known biological examples with sufficient evidence to convince one of an exclusively natural origin so it should not be a surprise that it pretty much does so. That doesn't make it a marker for intelligent causation, other than the intelligence behind placing the goal post that far away. Topic: The Algorithm Room William A. Dembski's writings claim that algorithms cannot produce Complex Specified Information (CSI), but intelligent agents can. A recent posting of Dembski's introduced qualifiers to CSI, so that we now have the "appearance of CSI" (or 'apparent CSI') and "actual CSI". Dembski categorizes as "apparent CSI" those solutions which meet the formerly given criteria of CSI, but which are produced via evolutionary computation. This is contrasted with "actual CSI", in which a solution meets the CSI criteria and which an intelligent agent produces. See and follow the link for "Explaining Specified Complexity". Dembski is also fond of both practical and hypothetical illustrations to make his points. I'd like to propose a hypothetical illustration to explore the utility of the "apparent CSI"/"actual CSI" split. Let's say that we have an intelligent agent in a room. The room is equipped with all sorts of computers and tomes on algorithms, including the complete works of Knuth. We'll call this the "Algorithm Room". We pass a problem whose solution would meet the criteria of CSI into the room (say, a 100-city tour of the TSP or perhaps the Zeller congruence applied to many dates). Enough time passes that our intelligent agent could work the problem posed from first principles by hand without recourse to references or other resources. The correct solution is passed out of the room, with a statement from the intelligent agent that no computational or reference assistance was utilized. Under those circumstances, we pay our intelligent agent at a high consultant rate. But if our intelligent agent simply used the references or computers, he would get paid at the lowly computer operator rate. We suspect that our intelligent agent not only utilized the references or computers to accomplish the task, but that he also used the time thus freed up to do some light reading, like "Once is Not Enough". There are four broad categories of possible explanation of the solution that was passed back out of the "Algorithm Room". First, our intelligent agent might have employed chance, throwing dice to come up with the solution, and then waiting an appropriate period to pass the solution out. Given that the solution actually did solve the problem passed in, we can be highly confident that this category of explanation is not the actual one. Second, our intelligent agent might have ignored every resource of the "Algorithm Room" and spent the entire time working out the solution from the basic information provided with the problem (distances between cities or dates in question). Third, our intelligent agent might have gone so far as to look up and apply, via pencil and paper, some appropriate algorithm taken from one of the reference books. In this case, the sole novel intelligent action on our agent's part was looking up the algorithm. Essentially, our agent utilized himself as a computer. Fourth, our intelligent agent might simply have fed the basic data into one of the computers and run an algorithm to pop out the needed solution. Again, the intelligent agent's deployment of intelligence stopped well short of being applied to produce the actual solution to the problem at hand. Because we suspect cheating, we wish to distinguish between a solution that is the result of the third or fourth categories of action, and a solution that is the result of the second category of action of our intelligent agent. We have only the attributes of the provided solution to the problem to go upon. Can we make a determination as to whether cheating happened or not? Dembski's article, "Explaining Specified Complexity", critiques a specific evolutionary algorithm. Dembski does not dispute that the solution represents CSI, but categorizes the result as having the appearance of CSI because the specific algorithm critiqued must necessarily produce it. Dembski then claims that this same critique applies to all evolutionary algorithms, and Dembski includes natural selection within that category. The question all this poses is whether Dembski's analytical processes bearing upon CSI can, in the absence of further information from inside the "Algorithm Room", decide whether the solution received was actually the work of the intelligent agent (and thus "actual CSI") or the product of an algorithm falsely claimed to be the work of the intelligent agent (and thus "apparent CSI")? If Dembski's analytical techniques cannot resolve the issue of possible cheating in the "Algorithm Room", how does he hope to resolve the issue of whether certain features of biology are necessarily the work of an intelligent agent or agents? If DI cannot distinguish between CSI directly produced by action of an intelligent agent and CSI produced by an intelligent agent one or an infinite number of steps removed, how can DI be regarded as anything except a Deist apologetic? If Dembski has no solution to this dilemma, the Design Inference is dead. Wesley R. Elsberry Wildlife and Fisheries Sciences Texas A&M University Graduate student and general bioacoustics dogsbody.