Uncommon Descent Serving The Intelligent Design Community

UK Guardian: Most of the next generation of medical and science students could well be creationists


ID is not the same as Creationism, however, it would be naive to say the following article has no bearing on the future of ID.

I’m personally disappointed to hear some creationist students mingling religious ideas into their scientific views, but on the whole, this report can’t be happy news for Richard Dawkins. :=)

Academics fight rise of creationism at universities

Most of the next generation of medical and science students could well be creationists, according to a biology teacher at a leading London sixth-form college. “The vast majority of my students now believe in creationism,” she said, “and these are thinking young people who are able and articulate and not at the dim end at all. …. Many …were intending to become pharmacists, doctors, geneticists and neuro-scientists.

[...] UK Guardian: Most of the next generation of medical and science students could well be creationists [...] A role model for ID-sympathetic college students, author and physician David A. Cook, MD | Uncommon Descent

Dave, since all 1000-bit coin sequences conform to your specification, it follows that all 1000-bit coin sequences imply design. Is that your position?

No, they don't all conform and you're become quite boring. Consider yourself warned. -ds secondclass

Dave, aren't all coin-flipping results equally unexpected?

If the coin is random, yes. The sequence you describe is statistically unexpected from a random coin and the improbability of it exceeds the universal probability bound. -ds secondclass

Dave, so I suppose that if I reported flipping a coin a thousand times with a result of all heads, you would not infer design unless I had a motive for cheating. Is this correct?

(Sorry that I'm beating this to death. I'm just trying to determine the extent of our disagreement. Thanks for bearing with me.)

I'd wonder what was causing nothing but heads to come up. -ds

Upon further reflection I think "statistically unexpected results from a well characterized physical process" is an independently given specification that fits this situation. -ds secondclass
Cellular Biologist Albert Voie, whom I mentioned at Another Pro-ID Paper Passes Peer Review writes at ARN:
The problem of calculating the probability for the emergence of a new functional protein in evolution is that we do not know the size of function-space. Function-space is less than sequence space, but we do not know how much less. If it is much lesser, the estimated upper limit for exploration of sequence space over the history of time (ref Gary) do not support blind search. But we do know something. We know that the key and lock systems between proteins and between proteins and ribozymes/DNA must be: A) "for a high degree of specificity the contact or combining spots on the two particles must be multiple and weak." An array of many weak interactions, such that all are needed to provide the necessary stability, will form a specific site for interaction. If only a few very strong interactions are used, there is an increased chance that a protein will find a similar interaction with improper proteins. B), "one particle must have a geometrical arrangement which is complementary to the arrangement on the other." In other words, the shape of the interacting surfaces must form a good fit, and this fit must be different from that with other proteins. Co-evolution of one key and lock system to another key and lock system is prohibited by the fact that mutations do not occur in parallel, both in the key and the lock. Therefore, and as shown by various studies point mutation in these domains impairs the system. This is also supported by studying the sequences of these proteins over a variety of different organisms where variation is zero or minimal. The probability of a key and lock system is thus general for all systems depending upon the complexity involved in the domain. The probability of a key with 25 amino-acids sites involved in the domain is thus ~ 1/20^(25) = 1/3.35e+32. In addition come the other domains that says something about what kind of a task this system really does. The suggested probability is thus a lower limit.
Oh well. I guess there is no future in pursuing this. Thanks for participating. Mark Frank

Mark Frank asked,

"You still have not given me a basis for choosing a specification when I have several to choose from (which is pretty much always the case in nature). "

That's because I essentially pointed out you can try them all. Try them all, assuming each specificaton has a sufficient number of bits. The link secondclass gave will give you the background you need without buying the book.

The link you provided from ARN, which you base your ideas had this to say:

" A full exposition of specification requires a detailed answer to this question. Unfortunately, such an exposition is beyond the scope of this paper. .... Although these conditions make good intuitive sense, they are not easily formalized. For the details refer to my monograph The Design Inference."

I pointed out the most tractable specification to apply where bits can be arrived at and where bits are effectively used in practice today is the specificity of protein lock-and-key systems. It is not too far from the question : "if a login requires a password with 500 bits, how many bits does the passwords' specification have? " Answer : 500 bits.

A lock and key protein system is analogous to the login/password system. scordova

Dave, so if I told you that I flipped a coin 41 times and got 40 heads and 1 tails, you would not infer design unless you knew of some advantange offered by flipping heads?

Correct! :mrgreen: If you bet me beforehand that you could do that and then did it you bet I'd infer design. -ds secondclass


You *must* calculate the probability of conforming to the template - otherwise the probability you have calculated is an unspecified probability. Suppose I draw a large target on the wall so the archer has a 50% chance of hitting it by accident. There are a hundred bricks within the target area. The archer shoots and duly lands in the target and, of course, hits one of the bricks. I then announce that the chances of the archer hitting that particular brick by accident were 1 in 200 so he must have been intending to hit it. I think you can agree that would be rather misleading?

If you don't believe me - ask Dembski. I am sure he will concur with the statement "to calculate the complexity you need to estimate the probability of the event conforming to the specification"


PS I beginning to wonder if I shouldn't be answering the questions - not asking them :-)

Your analogy breaks down where you switch from a target of 100 bricks to a target of 1 brick. You definitely shouldn't be answering questions and I'm beginning to wonder if you should be asking your questions on another blog. You have come here with a preconceived notion that you've discovered a fatal flaw in making design inferences and won't take no for an answer regardless of the fact that design inferences using the same procedures are routinely used in forensics, fraud, cryptanalysis, SETI, paleontology, and anywhere that chance vs. design is relevant. -ds Mark Frank

A few comments directed at various folks...

To all bloggers: I'm still confused by the specification that DaveScot attributes to the Caputo sequence. If there were no advantage to being first on the ballot, would the sequence be unspecified? What if the lone R were a D? What if it were an isomorphic bit string?

To Mark Frank: In Dembski's definition of specificity here, σ = –log2[phi(T)·P(T|H)], the latter T presumably refers to a composite of all events that conform to the given pattern. Therefore, a more general specification would tend to result in lower specificity. To determine whether design can be inferred, one would presumably choose the specification that gives us the highest specificity. This would entail a description that is both narrow and simple.

To DaveScot: It seems to me that the chance calculation can be meaningful only if it corresponds to the chosen spec. If the spec is "aids survival," then the chance calculation should give us the probability of any event that aids survival. Likewise, if the spec is "clotting mechanism," then the chance calculation should be appropriate to that spec.

If there were no advantage to being first on the ballot, would the sequence be unspecified?


What if the lone R were a D?

Still specified.

What if it were an isomorphic bit string?

I'm not sure what you mean by that.

Specifications only serve to separate meaningless random events from events that are possible cases of design. You're reading way too much into specifications.

For instance take the string "kriemvpoaignkodaaarwf". If we know nothing about it it's just a random string of numbers and letters with a probability of 1 in 1144561273430837494885949696427 of appearing by random chance. Is it designed? We don't know. Random chance generates highly improbable patterns all the time. If we find out it's the combination to a lock then we have a design inference. No say the pattern is 114 which has only a 1 in 1000 chance of occuring randomly. If it turns out to be a lock combination we can't make a design inference because 1 in 1000 isn't sufficiently improbable. Specifications merely separate meaningless patterns from meaningful patterns. -ds secondclass

Oh dear - we are going round in circles. If the specification is simply "aids survival" then the calculation you need is not the chance of forming a clotting mechanism - it is the chance of forming something that aids survival - a very common occurence - it happens every time a bacterium evolves resistance to a drug. That's what I mean when I say the precise wording of the specification affects the complexity dramatically.

That is wrong. The chance calculation is the formation of the clotting mechanism not the formation of anything that aids survival. -ds Mark Frank

Dave thanks for a clear and productive response.

Not surprisingly it raises some interesting follow-on questions. The choice of specification will affect the probability of the observed outcome conforming to that specification so if any independent pattern will do how are we to handle the case where on person has chosen an independent pattern with low probability and another chosen a pattern with high probability?

Take the SETI example, suppose I am studying radio signals and one day receive a stream that can be digitized and turns out to be the binary for the first 10 prime numbers. I could use the specification "generates the binary for the first 10 prime numbers". I could then make some calculations of the chances of this happening in a random sequence of digits. The answer would be extraordinarily low. However, another independent specification might be "generates the binary for any recognised sequence of ten numbers" (the first ten squares, the first ten in the fibonnaci series etc etc). This probability is considerably higher. In fact it would be surprising if this did not happen from time to time if you were watching a random sequence of bits day after day. Using the first specification one might invoke the design inference. Using the second sequence the case for invoking the design inference is much weaker. Which is the right one to use? Why?

I am sorry that my previous response to Salvador was rather self-defensive. I really want to keep this polite and constructive. If this is well-worn terroritory with established responses then give me the reference and I will stop.


PS Do you get up extraordinarily early or are you based somewhere other than the USA?

The choice of specification will affect the probability of the observed outcome conforming to that specification

Blood clotting is something that aids survival. It conforms to the specification by definition. If it didn't conform it wouldn't be a specification and we couldn't make a design inference. What we don't know is the probability of chance forming the clotting mechanism. Whether specified or not won't change that probability. If the probability is above what Demski calls the "universal probablility bound" of 1 in 10^150 and it is specified then we have warrant to infer design. Specification in biology is only a matter of identifying a practical function the pattern of interest serves. -ds

Mark Frank


The URL I was referring to was http://www.arn.org/docs/dembski/wd_idtheory.htm which I mentioned in #29 above. As I said there are probabilities associated with outcomes matching patterns - but you were talking about the information content of specifications - not the outcomes! You even confused secondclass and then you corrected yourself 35. I humbly submit that you were the one who was phrasing things incorrectly. Incidentally which ideas am I phrasing incorrectly? You don't say.

You still have not given me a basis for choosing a specification when I have several to choose from (which is pretty much always the case in nature). You seem to waver between "it doesn't matter - choose whichever one you like" and "choose one that gives a high complexity/information content for the observed outcome". Does NFL have the answer to that question? Is it so very complex that I need to read a book to get the answer?

(NFL is not available in the UK - I can only get it at a large price and long delay from the USA - so I need a pretty strong justification)


The pattern you used "something that helps an organism survive" is fine. The point of a specification is merely to distinguish signal from noise. Any independent pattern will do. In biology, specification will usually simply be function or purpose. You defined a purpose and the blood clotting mechanism fits that purpose. Many moons ago someone here kept insisting that we run junk DNA through the design filter to see if was really junk or not. I kept telling the guy, but he didn't understand, that junk DNA has no specification. Until we know something has a purpose we can't make a design inference. Just so, someone here (possibly you) offered me the string DDDDDDDDDDDDDDDDDRDDDDDDDDDDDDDDD and asked me what the specification was. I had no clue. Then he gave me link where it was explained that D and R have specific meaning and more of one than the other represents a survival advantage (for politicians in elections). Thus it became specified but only when my background information was sufficient to see a detached pattern that the string conformed to.

Finding complex specified information isn't the end of a design inference. It's just the beginning of the hardest part and that is determining if there are sufficient probalistic resources to generate the CSI without intelligent agency. In other words, assessment of chance hypotheses. Random mutation + natural selection is a chance hypotheses and in fact it's the only known chance hypothesis for biological complexity which is why it gets all the criticism targeted at it from ID proponents. -ds

Mark Frank
Salvador, thank you for your clarification. That makes sense. secondclass


The ULR for mark_frank.blogspot.com does not work so I have no access to the papers you are talking about. You say you've read the relevant papers, what books by Dembski have you read?

"Dembski has provided a clear definition of information in the paper I referred to and it is clear that patterns are not the kind of thing that can hold information because there is no such thing as the probability of a pattern."

Please give the definition or the name of the paper you are referring to. Page 78 of NFL indicates complexities associated with specifications. There are probabilities associated with outcomes matching an independent pattern. You're not even phrasing the ideas properly. Please list the books by Dembski you have available to continue this discussion, otherwise we have no common base to argue what he actually said.




The following link gives highly shortened tutorial of the concepts:

When I say a 500-bit specification, it is a shorthand for a specification that would identify an event or set of events that have 500 bits of information.

The specification must create a region within the space of possible outcomes. That specification has a number of bits associated with it. I made a simplification of saying each specification must be 500 bits.

To be excruciatingly formal, we would actually take the union of all specifications we can reasonably enumerate, and that set itself will form a specification of slightly less than 500 bits.

For example, the union of a million 500-bit disjoint specifications would result in a composite specification of about

500 - log2(million) = 480 bits

Given we can't explicitly list more specifications than the number of atoms in the universe, the largest possible union of 500-bit specification would be around:

500 - log2( 10^60 ) = 300 bits which is a very tiny target even though it has as many specifications as there are atoms in the known universe.

One could of course require the individual specifications be 700 bits instead of 500, and then, the net result is the most general composite specification could not be more than 500 bits. But I've simplified some of the considerations above.

Everyone, we've gone off topic, but that's ok, since I think all the commentary on the original has been said already, and this would be a good place as any to talk about the new issues while I have people's attention.

Salvador I am sorry if you think I am being argumentative for the sake of it. I am not. I believe that unless this question is answered then there is a major flaw in the design inference. If that is true, then surely it is important and not just an undergraduate quarrel. And I have read quite a bit of the relevant papers. Dembski has provided a clear definition of information in the paper I referred to and it is clear that patterns are not the kind of thing that can hold information because there is no such thing as the probability of a pattern. The only things that can hold information are outcomes which may or may not conform to patterns. So information content cannot guide me in choosing patterns. I think you may be getting confused with patterns that turn out to be hard for outcomes to match i.e. only outcomes with high information content match them. But that cannot be our basis for choosing patterns because the whole point is that the pattern should detached and independent of the probability of the outcome. You say matching any pattern is sufficient. OK then I will pick a pattern for blood clotting - blood clotting clearly conforms to the pattern that it is a system that helps organisms live longer. What are the chances of a species randomly evolving something that helps it live longer? Very high indeed - we see it happen in microevolution all the time - ergo the complexity of blood clotting is low - ergo there is no design argument. To be a bit more down to earth - you can wonder at the sequence of chances that led to a particular set of proteins combining to clot blood. By itself it may be appear incredibly improbable (I have my doubts - but I will assume it). But this is to ignore the other paths that evolution might have blundered down and come up with a workable solution. It might, for example, have come up with hardened blood vessels that rarely punctured. So if you change the question to - how likely is it that evolution would come up with an answer to the problem of bleediing then the answer might actually be much higher and is certainly very hard to calculate. How do you decide which is the right question? What justifies the decision to only look at that probability of evolution coming up with particular combination of proteins? Anyhow I will leave this now. I have asked the question and I believe no one has been able to provide an answer. I suspect further discussion would not be productive. Thanks for trying. Mark Frank
Salvador, my understanding is that the event must be at least 500 bits in order to meet the complexity threshold. I don't understand your reference to 500-bit specifications. secondclass
Mark wrote: "The problem is how to choose between them. I don’t believe information content is much help." Then you believe wrong. You don't need to choose between them. Matching any single one of them is sufficient. One cannot possibly list all the 500 bit specificaitons possible, that is why even having a trillion trillion of 500-bit specs to choose from does not substantially increase the probabilty the pattern match was due to some post dictive fluke. You're being argumentative, and I'd appreciate if you'd take time to study the material you presume to criticize. Salvador scordova
I see your concern. But just as the picking the best Darwinian pathway isn’t a function of an equation with a rote answer neither is choosing the most apt independently given pattern. -ds But I can give criteria for choosing the best pathway. It is the one that is most probable given what we know. It may be hard to assess which one is best but we know the kind of argument that we can use in favour of one over another. You don't give any reason for preferring one specification over another. They are all functional. None of them have probability or information content - that's not the kind of attribute a pattern can have. So far it seems to be completely arbitrary - which surely is not right? Mark Frank

First thanks for all your answers.

Patrick - I haven't bought the books but I have read several of Dembski's papers on the Internet and none of them deal with the question - how do I choose which specification? I might well buy one of the books shortly (especially if you can tell me which one contains the answer!) - but surely it is not such a difficult question that one of you can give me the answer?

Salvador - you say there are multiple specifications - which was precisely my point. The problem is how to choose between them. I don't believe information content is much help. How can a template have information content as Dembski defines information as a function of probability (see http://www.arn.org/docs/dembski/wd_idtheory.htm)and templates don't have a probability?

Dave - thanks you have at least told me what the template is for blood clotting. Does everyone else agree? However, you didn't answer the second vital part of the request. How did you choose that one as opposed to the many other templates that blood clotting conforms to? Or are your last two sentences meant to answer that question?

Specifications in biology are usually function or utility. It was chosen as a good example of irreducible complexity

The first sentence clearly doesn't answer the question, as all the alternatives I proposed were to do with function or utility. The second sentence is almost as incoherent as saying specifications have a probability. It is outcomes that are irreducibly complex (if anything is) not specifications.

I am sorry - but I still don't see an answer that makes sense. I hope you can see my concern?

I see your concern. But just as the picking the best Darwinian pathway isn't a function of an equation with a rote answer neither is choosing the most apt independently given pattern. -ds Mark Frank
Mark Frank, I already pointed out there are multiple-specifications and one need only identify a component for a rigorous analysis such as a protein lock-and-key to infer design. You're insisting on spec, and I've already pointed out that is invalid, yet you ignored what I said. Salvador scordova

But specifications aren't alternative hypotheses like evolutionary pathways. They are patterns. It makes no sense to to talk of one being more likely than another. Look at the examples of specifications above for blood clotting. The blood clotting system meets them all. You can't say that the specification "a system that clots blood" has a probability or likelihood - it is just one of many patterns that the actual system happens to conform to.

I am surprised that in this forum - the very centre of the ID movement - I can't get a coherent answer to such a basic question. Maybe someone else knows the answer? Or at very least tell me what the specification is for the blood clotting system and why it has been selected :-)

Blood clotting conforms to the independently given pattern of patches or barriers - a class of structure that stops the loss of critical fluids (or gases) and/or forms a barrier against the entry of contaminants. There are analogs in plants, many kinds of human engineered machinery, and even artificial substitutes like bandages or holding your hand against a wound. Was it really that hard? Specifications in biology are usually function or utility. It was chosen as a good example of irreducible complexity. -ds

Mark Frank

Mark Frank,

There are multiple specifications which can be used as templates to detect design. An important requirement is the specification have sufficient information content (500 bits is customary, but not necessary). The answer to your question is you can use one, two, any and all the specications provided each specificaiton is sufficiently complex.


Mark Frank, it's probably best if you purchase Dembski's books on the subject. They're not that expensive and they'll definitely do a better job of explaining ID than the short responses you will receive in the comments section. Patrick


I would examine the relative pathways and choose the one that seems most likely. So will you choose the specification that gives the greatest chance of being actualised i.e. the lowest complexity?


I would examine the relative patterns and choose the one that seems most likely. -ds Mark Frank

DS - I can differentiate between hypothetical Darwinian pathways. But you still haven't told me how you decide which is the right specification to use. What makes one specification more appropriate than another?

I think it's the same way you know which hypothetical Darwinian pathway to use. Tell me how you do that then I'll tell you if it's the same way to choose between specifications in ID. -ds Mark Frank


Thanks for replying. I guess the blogspot web site must have been down. However, it is working now and Dave Scot has kindly replied using that site. I don't want to clog up this blog with my questions so I have responded there: mark_frank.blogspot.com

(In essence my response is - all the specifications I offer for blood clotting meet both Dave and your criteria. My concern is how you choose between them given the large difference it makes to the complexity. It seems to me there has to be a principle for doing this or ID has a major problem.)


It's the same principle we use to discriminate between hypothetical Darwinian pathways in the evolution of blood clotting. You know what that is, right? ;-) -ds Mark Frank
Mark Frank, Welcome to our weblog and thank you for reading. I tried clicking on the links you provided, and I was unsucessful in getting to the websites. My apologies if there was something there you were hoping for me to comment on. 1. Regarding your first questions, really, anything is subject to investigation for design. Somethings, however, are much amenable to rigorous mathematical analysis, some are not. Seeing the design in a statue of pharoah is quite easy, but making a rigourous mathematical case that it's designed is far more challenging. I think ID is in it's infancy as far as finding objects which we can subject to rigourous mathematics. The blood clotting cascade is a great candidate, but we might find it easier to compute the design content of certain parts of the system, versus the entire system. For example, the protein lock-and-key pairs within the system are amenable to mathematical estimates. 2. Regarding the problem of where specificaitons come from. If I may offer a radical answer, which many IDists may not feel comfortable with: "the specifications come from existing human engineered systems." In fact, unless the systems can be demonstrated to be analogous to something a human-like intelligence would make, we would not be able to recognize the system as designed! Salvador Cordova scordova

physicist, I agree with you, as well as valerie, but I also think that there is a distinct prejudice in science against evidence that indincates purposeful structuring in nature.

I see this as a counter-fanatical imbalance, and I can prove that this political aspect of science creates willful ignorance of any evidence that suggests that we're not here by accident. It has been my honest experience that the mentioned "counter-fanatical prejudice" affects many scientists willingness... to follow the evidence where it leads, (to borrow valerie's coined phrase)... even if it contradicts a closely-held belief and upsets their fellow believers.

There is no doubt in my mind, (from much experience with this as a previously naive and honest scientist), that this puts an honest scientist at an unfair disadvantage when attempting to offer evidence for first principles that are open to closed-minded interpretations.

What you "think" that you *know* can be completely turned around to mean exactly the opposite of what you think that it means, given one new missing piece of the puzzle, so be very careful with what you think that we really "know", for as long as we have no ToE.



Leave a Reply