I am working on a series of tutorials to cover the basics of Intelligent Design, especially the mathematics of it. This is my tutorial on Specified Complexity, and I would appreciate any thoughtful criticism of it.
Note that I am specifically requesting criticisms on the content of the video itself, not on applications of the concept that are outside the bounds of the video. This is both to help me (I’m trying to improve my presentation of ID) and to help clarify the conversation (is the criticism of *this* information or of some *other* information).
CSI is nonsense.
http://www.talkreason.org/articles/eandsdembski.pdf
If IDers want to accomplish something, they need to do some science. Year after year of blogging isn’t getting anywhere.
AhmedKiaan,
I’ll talk about this paper in a later post (it’s not that great of a paper), but by posting it it seems you aren’t actually watching my video or responding to it, since the paper is mostly responding to other claims than the ones presented in the video.
Skimming through your presentation, I am puzzled by one thing. Does “chance” include natural processes such as natural selection? Or does it include only random processes such as random mutation (say, monkeys typing on a 4-letter typewriter)?
If the former, then calculating P(T|H) will be very difficult in cases where biological evolution is involved. If the latter, then I will argue that seeing CSI does not rule out that the adaptation could come about by natural selection.
This distinction is, as far as I can see, the difference between Dembski’s use of CSI to detect design pre-2005, and his use after 2005. He started out with P(T|H) not incorporating natural selection, then he changed to having it incorporate it. The former case used a conservation law the Law of Conservation of Complex Specified Information (LCCSI). I have argued in 2007 that that law does not do the job. The useage after 2005 incorporated natural selection. That made the whole design inference useless, since we have to rule out natural selection first, in order to do a design inference whose object is … to rule out natural processes including natural selection.
So the crucial issue is P(T|H). Does it include natural selection?
This seems to be the version of CSI where highly compressible data is taken to posses lots of information. That’s weird to say the least. Which version is it?
Joe Felsenstein,
As I noted in the talk, P(T|H) can include whatever you want, depending on what you are trying to show. I should note that the goal of this talk is not to show that any particular thing is designed or not, but how the logic of how CSI works in general. It *can* be applied to biology, but I don’t really want to get into the weeds of that at the moment.
But, to answer your question, if the goal of using CSI was to deal with the origin of life, then P(T|H) would not include natural selection. If the goal of using CSI was to deal with subsequent evolution, then, depending on what you are trying to show, it should include natural selection.
CSI merely provides a way of finding “surprising” results where the bare probability distribution itself wouldn’t be able to denote something as such. “Surprising” will depend entirely on what you include in P(T|H). Why you would want to know if something is surprising will determine what you include in P(T|H).
dazz,
I use the term “specified complexity” in the video instead of CSI because I think that it is a better description, for the reasons you indicate. That being said, I think that one could argue for some types of CSI being information, if the compression is not based on descriptions with a low logical depth. I don’t know how one would compute a number for this, however, but it is a worthwhile area to think in.
Jon, you are asking a lot in hoping folks here are going to sit through a 45 minute video. On the other hand, I did manage to flick through it to read the text summaries you provide. Have I missed much?
BTW talking about design as an alternative to some process such as biological evolutionary change seems entirely wrong. Designs always require a designer. In the case of evolution, the niche environment fulfills that rôle.
Alan Fox,
I agree that it is a lot to ask, but my goal is to (a) improve my video (which would require watching it) and (b) talk about specific questions on specified complexity (which would also require watching it), as opposed to numerous other peripheral questions about it. The video walks through a very simple example with the goal of removing ambiguities. and being very specific about what it is that is or isn’t being tested/discovered. Most criticisms of specified complexity get bogged down in whether or not X is accounted for, whether it really means Y and so forth. My goal here is to just get at the mathematics, and whether or not the bare mathematics are sound, irrespective of whether a particular application is sound.
Note that using “specified complexity” as any kind of design inference would lean more on the “particular application” side, though I do present a few important points about doing so in the video which I would be happy to discuss.
Nobody in science uses CSI. It’s just good for fooling laypeople who want to believe it.
If someone wants a super-short summary, basically I am showing how specified complexity can be used to reject a hypothesis where all outcomes are equally probabilistically small (or adjusted to be so by P(T|H)), by providing a way of measuring how “surprising” the sequence is even when its bare probability is equivalent to another. That is, to use the example in the video, if you told me you flipped heads 1,000 times, for what reason should I not believe you?
Also, by the way, the discussion given is based on Ewert (from here, paper available here) more than Dembski, with Dembski’s notions being given as an extension of Ewert’s.
https://en.wikipedia.org/wiki/Statistical_hypothesis_testing
I’ve watched the presentation until the 30 min mark. This one left me confused:
Specified complexity was developed to detect design
Design, in this case, is a purposeful act of will
– As an act of will, it is not dictated by a predefined law.
– As a purposeful act, it is not haphazard
(…)
As an engineer I always thought of myself as a designer, but according to your definition I might have or I might not have created any design. So, do we have examples of actual designers creating design (besides Jesus)?
This is crucial. Otherwise it’s just the Sherlock Holmes argument. Until you are sure you can say you have all possible outcomes covered, the default argument is not justified. And how you can be sure when you have all outcomes covered in any situation more complex than coin-tossing and the like, well, it beats me!
PopoHummel,
Can you be more clear as to how you are confused? I thought it was pretty clear, but perhaps I am mistaken.
I would argue that:
– When PopoHummel engineers it is an act of will, it is not dictated by a predefined law.
– When PopoHummel engineers it is a purposeful act, it is not haphazard
Would that be correct or incorrect? Thus, by the definition, I would say that when PopoHummel engineers is a purposeful act of will, and therefore design.
Alan Fox,
The default argument, actually, is happenstance, or no result.
Now, as a more theoretical distinction, if you wanted to be pedantic about results in general, every single statistical argument, ever, relies on this to some extent. It may be the fact that this argument relies moreso on this, but literally every statistical argument for anything assumes that there is not some alternate axiom that you aren’t aware of that interferes. If your goal is to never be wrong by unknowns, you shouldn’t be doing statistics or science.
It is also true that if there were an unknown law in effect, you would get positive specified complexity. However, I think that Dembski was assuming that if someone found a regularity that was indeed law-like, that they would do the rational thing and propose a new law. However, Dembski himself doesn’t give a criterion for law-like vs. non-law-like compressions. This created additional confusion as Abel and Trevors (2005) claimed that functional information was not compressible. Ewert (2014) introduced the idea of compression through functional tests, which, as I noted in the video, can be used to distinguish the types of regularities. If the compression has a large logical depth or refers to system-wide teleology, then it is a design-oriented compression. If it has low logical depth or refers only to localized regularities, then it is a law-oriented compression. The shorter one wins.
I’m sure that these considerations aren’t perfect, but I do think that they move the discussion forward.
Commenting as a mathematician and probability theorist — the whole notion of CSI is nonsense.
In your “1000 heads” example, you talk of “physical complexity”. But it is really “logical complexity”.
You say that you can describe the 1000 heads in 64 bit, rather than 1000 bits.
But this is completely wrong. What you can really says, is that you can describe it in — 64 bits of information together with millions of bits of assumed human conventions.
So it really isn’t just 64 bits. And Kolmogorov complexity is itself a dubious concept, because it depends on a huge volume of assumed human conventions. Kolmogorov recognized this, in that he used his notion of complexity as an asymptotic measure rather than as a specific measure of small things.
CSI, however, is not an asymptotic measure. It pretends, falsely, to be an actual measure of small finite things. It is nonsense.
I don’t have time to go through the video right now but I know that one of the big criticisms of CSI is that nobody has ever attempted to use it to distinguish between something that is known to be designed (e.g., a fishing reel) and something that is not (e.g., a snowflake). Do you have any practical examples using real world designs in your presentation?
If CSI could be calculated or used there would be a list of things and their CSI values. There is not.
A tutorial that tries to explain the concept apart from any application of it? This is against the very foundations of didactics. Things are learned better by specific examples, always.
Just my two cents. I didn’t watch further than two minutes.
Neil Rickert,
So, it is definitely true that Kolmogorov complexity measures have an implicit or explicit +C at the end, meaning that it is more meaningful in the limit than for small, finite sets. But pretty much the entirety of science is lopping off things that are true in the limit and using them as provisional truths for active inquiry. As an example, the identity of indiscernibles is used in chemistry. However, it is not possible to actually exhaustively list the properties of substances. Thus, while it is only technically true in the limit, we provisionally accept it for the finite circumstance.
But, nonetheless, the interesting thing about the Church-Turing thesis is that it basically establishes a core meaning to what finite logical machines actually do. They are basically all equivalent. Now, if you think that some feature of a KC description invalidly uses some extravagant feature of a particular programming language, that’s fair enough. But I think on the whole using extant programming languages actually overstates KC by quite a bit, because (a) the whole ASCII character set is not needed, and (b) a lot of sequence space is taken up by syntax errors. Really my KC should be much lower than 64 bits, as 64 bits *might* be the lowest number of bits to produce a 1,000-bit sequence to begin with.
However, what makes KC quite useful is that it provides a way of providing independent descriptions. If that makes it off by an unknown constant factor, or means that my description might be oversized because of language limitations, I think that can be cleaned up on the backend (with further research on programming languages and their relation to mathematical descriptions), or taken, like most of science, as a finite estimator.
The argument, as I understand you, is that it converts an impractical mathematical truism to a practical but imperfect tool. Yes, it does.
I am confused, because given your definition, it is unknown whether a human can create design, since it is unknown whether our acts are based on will or predefined law.
I don’t know what the underlying process is that enables me to engineer. Is it will? Is it predefined law? No one knows (except for Jesus).
What Neil said. CSI is garbage. It’s useless.
PopoHummel,
If our acts are based on predefined law, then having any argument is the epitome of ridiculousness. I literally couldn’t do anything but write what I am writing, and you literally could not do anything but write what you are writing. So, the alternatives are, (a) life is a ridiculous theater where everyone is fooling themselves all of the time, or (b) we are able to make choices. (a) might be true, but if (a) is true then everything else (and I mean everything) loses its justification.
JB, that is just the same lame argument used by WJM, Gordon Mullings and others at UD to argue against materialism. How can a bag of chemicals make a decision? For some strange reason they think that it is a knockout argument against evolution and materialism. Even stranger, I don’t think that they understand why it is a lame argument. As Mullings would say, that speaks volumes.
But getting back to your presentation, I suggest that you provide some real. practical examples. What is the specified complexity value for an ABU 300 fishing reel? What is the SC value of a string and cane fishing pole? What is the SV value of a stone tool? What is the SC value of a snowflake? What is the SC value of the Amazon river? If you can include these values in your presentation, clearly showing that the approach can distinguish between things that are clearly designed and those that aren’t, you might have the beginnings of a valid argument for CSI and ID. But only the beginning because the inference will still be based on the extrapolation from a single designer to his artifacts. In stats we call that zero degrees of freedom.
You could improve your statistical power by using the same tool to calculate the SC value for non-human designs. For example, show that it can distinguish between a beaver dam (designed) and a natural dam caused by wind and water movement. Or distinguish between the designed bird’s nest and a cluster of twigs lodged in the branches of a tree.
No, that’s wrong. That’s a complete misunderstanding of science.
Yes, it is true that scientific theories idealize to some extent. But there is a strong pragmatic core to science. There is no comparable pragmatic core to specified complexity. That’s why you cannot actually tell us the amount of specified complexity in a banana or in a shoe string.
Only in the sense that it is true by definition as a principle of abstract logic. It does not tell us anything about real life.
That’s what the true believers say. But they are not independent of human conventions.
No, not at all.
The argument from me is that it falsely claims to convert an impractical mathematical truism to a practical but imperfect tool. But that’s plainly wrong. It has failed to provide that alleged “practical but imperfect tool”. It is, at most, a tool for pushing religious ideology under a pretense that it is science.
I’d reming you of a saying Kairosfocus is fond of repeating; “Every tub must stand on its own bottom”. Statistics is a great tool for testing hypotheses – seeing if the data (from observation, experiment etc) is a good fit for the explanation. To fit data to a hypothesis, first you need a hypothesis.
Isn’t that a whole lot of effort to get around the problem that you don’t actually have any substantive evidence for design in life?
If we found a machine on Mars we’d have no problem making the case that intelligence made it. The rationality and probable (at least) purpose actually point toward intelligence and “design.” If we found life, well, get real…
What are the odds that we’d end up with the derivative patterns of life (nested hierarchies) if we were designed by any known intelligence, even if we increased that intelligence exponentially? What are the odds that the predictions of evolutionary theory (sans intelligence) would be matched by a very intelligent being?
I don’t really have a problem with CSI as a concept, although clearly the pretense of applying it to life ignores the fact that we don’t know the probabilities at this time. The bigger conceptual problem is that there are many probabilities, and the ones that strongly point toward evolution are generally ignored by IDists. Of course, if they didn’t, the fakery of calculating CSI to pretend that evolution can’t happen would be immediately exposed.
Glen Davidson
And, for bonus points, is it more or less the then the value for the ABU 507 model?
Obviously less. 300 is smaller than 507. Duh!
GlenDavidson,
This is not what the CSI argument is trying to do. It is not looking at the probability of the pattern forming. It is observing the pattern and determining if it was caused by a stochastic process. If you got 2 perfect bridge hands in a row the chance of it happening is 100%. The next issue is if the process was stochastic or a stacked deck of cards.
If you look at Joe Felsenstein’s argument he is making the point natural selection makes the mechanism non stochastic.
This is a valid complaint but when we observe the large micro machines of life it appears large stochastic leaps are required before selection can take hold. Alas we have the conference at the Royal Society a month ago.
Yeah, who cares about your pedantry? We know what the “argument” is, and that you, like other IDists, consistently ignore the probabilities that point toward unthinking evolution.
Really? Like that isn’t obvious. But we know what the sham of ID constanly claims.
Well of course it isn’t stochastic. Why are you stating the dull and obvious, except to ignore the fact that CSI is essentially worthless while the odds that we’d see the patterns we see from any known intelligence are astronomically against?
Well, too bad you have nothing but your hyperskepticism against the science to back that up.
Alas, you have no meaningful evidence for ID at all. Until then, there’s really nothing to ID.
Glen Davidson
Are you talking about the (in my view simplistic) idea of a probability threshold of 1 in 10^120? Pass the threshold and must be “design”. How does this determine anything?
johnnyb,
Hi Johnnyb
I think your tutorial is quite good but more for people who are very familiar with the sequential space arguments.
Can CSI help the design argument?
The jury is out, but I think getting more people on the same page as far a definitions go can help those interested come up with new arguments.
For me, I appreciate you making the tutorial. 🙂
GlenDavidson,
I am interested in your calculation of the odds against and intelligent cause.
I am interesting to see if you can form a real argument that CSI is worthless.
OK, so you are using the post-2005 Dembski definition, and when applied to post-origin-of-life evolution, one needs to somehow take account of how probable it is that natural selection comes up with the pattern, not just mutation.
As number of people have noted here, that leaves the heavy lifting to the calculation of P(T|H). If that is small enough, it is what does the job of calling into question whether natural evolutionary forces can do the job. All the CSI machinery is dependent on that and basically then adds nothing to the design inference.
I am interested in any actual evidence for an intelligent cause. You continually fail to provide any.
I don’t have to provide evidence against your worthless drivel at all. As for calculations, I’m afraid it’s up to you to show that any intelligence is likely to produce the patterns predicted by evolution. The bankruptcy of your ideas shows every time you demand of others what you fail to provide.
I am interested if you can understand anything in context. I already stated that I don’t have a problem with the concept but with its use, providing the context in which my brief point about the worthlessness of CSI should be understood.
Anyway, it’s up to you to show that CSI can show what you claim it can. Your burden tennis is tiresome, as you care not a whit that you haven’t the least evidence for your precious ID.
To repeat: Alas, you have no meaningful evidence for ID at all. Until then, there’s really nothing to ID.
Deal with it. Of course you won’t.
Glen Davidson
I disagree. The notion of specified complexity is due, not to William Dembski, but to Leslie Orgel. Essentially the same notion has been used in the paper by Hazen and Szostak (as “functional information”, and also by me in 1978 as “adaptive information”). In those discussions there was no argument that natural selection could not achieve a high enough value of SC to qualify as CSI.
Orgel, Hazen, and Szostak are not people to casually dismiss. What doesn’t work is Dembski’s additional Law of Conservation of Complex Specified Information, which was what he originally tried to use to show that natural selection could not put CSI into the genome. His argument didn’t work, but the original notion of SC is something you can calculate in simple evolutionary models — and in those models you can see it increasing as a result of natural selection.
Large amounts of SC are self-evidently present in living things, which are far better adapted in multiple ways than could ever be achieved by monkeys typing on 4-letter typewriters.
In the 2002 discussion in No Free Lunch, Dembski said that specified complexity could be “cashed out” in various ways. Only one of these was by finding low algorithmic information. Yes, I find the Kolmogorov Complexity argument weird too. It’s one thing to use fitness as a scale, and try to argue that natural selection somehow can’t get you to the high fitness. But the Kolmogorov Complexity seems just arbitrary in this usage.
A hummingbird has high fitness and lots of highly functional systems. A party ballon has a simpler description than the hummingbird. But it has a higher specified complexity? Why is that helpful to anyone?
I agree — I’m definitely not going to spend that much time on a video. A paper is a much better format because it puts the flow in the hands of your audience. It’s much easier to make notes, go back and forth, and really understand the arguments being made.
Videos strike me as similar to the debates creationists love to challenge scientists to have. The goal is persuasion and point scoring, not understanding and constructive discussion.
How about at least giving your audience both options?
GlenDavidson,
You continue your path of circular reasoning by making un supported claims. You made a claim of a lack of probability and then refuse to back it up. The evidence of design is invisible to you because of extreme cognitive bias.
When I see a biological system that looks and operates in a very similar way to a human design then the inference to design is immediate.
Your answer is that it is inadequate evidence but you have no standard to support your claim of inadequacy except your personal incredulity.
I think it is time to agree to disagree.
Way to project. It’s just a fact that the patterns of derivation is extremely improbable via any known process other than evolution. Your attempt to avoid that fact by demanding calculations is another indication that you can’t handle discussion of the evidence, let alone back up your vacuous claims.
I can readily note that shooting a target 500 yards away with one try without aiming is highly unlikely without doing the calculations. Your lack of dealing with the issue by playing burden tennis is appalling, if hardly unexpected.
Yes, so if you find an internal combustion engine made by aliens you don’t have to calculate whether or not it is due to intelligence. Life, of course, is in many ways quite unlike human designs, even if similar in some ways (how could it not be if life is functional? Physics won’t allow function to be wholly different).
Yet another claim you can’t back up. That makes, well, nearly all of them.
It’s not that simple, because your disagreement is based upon no sound evidence, while my disagreement is based heavily on the evidence. Evolution wouldn’t exist without evidence, ID constantly does.
If you’d even begin to deal with the issue of evidence with a modicum of openness, your position would be immediately destroyed. Hence the endless attempts to attack the only theory of biologic origins that resulted from close consideration of the evidence, heedless of the fact that your position has no evidence that can withstand scrutiny.
Glen Davidson
One of the many fatal flaws with the CSI nonsense is the people offering it as evidence of design never take into account the feedback inherent in evolutionary processes when doing their calculations.
The probability of being dealt a royal straight flush with five cards from a standard 52 card deck is 1 in 649,739.
If you are allowed to discard one card and redraw once the probability is higher.
If you are allowed to discard one card and redraw three times the probability is higher still.
If you allowed to discard and redraw as many times as you like the probability will be 1.0 i.e certainty.
What IDiots do is walk into a room where evolution has been discarding and drawing better genetic “hands” for over 3.5 billion years, then demanding the probability be calculated based on a one time snapshot of current results.
That’s what passes for higher math in IDiot circles.
I see. All you have is the standard argument from ignorance based personal incredulity i.e “it looks designed to me so it must be designed!!’
You’ll understand why people who study and work in the fields of the evolutionary sciences find that less than persuasive.
“You’ll understand why people who study and work in the fields of the evolutionary sciences find that less than persuasive.”
Doubt it 😀
Thanks Prof. Felsenstein
I’m afraid we all know that being helpful to scientists is not what Dembski’s shooting for
Somewhat indirectly, you raise a point that underlies many of these discussions. Creationists are masters at VERY carefully missing the point. You can keep talking about “natural selection” forever, and they NEVER actually see the word selection there.
Another standard is to posit some claim evolutionary biology has never made, and show that claim to be false. Douglas Axe has made a career of this. I could cite many other examples.
But the interesting question is whether creationists are making honest errors due to what theological blinders prevent them from seeing in the first place, or whether they are deliberately misrepresenting, quote mining, misdirecting, changing the subject etc. knowing their faith is false but unwilling to admit it.
In practice, ID represents the effort to find some rationalization, ANY rationalization, to support a false foregone conclusion lacking any evidence. Whether this is due to mendacity or stupidity, the sheer determination and tireless effort is remarkable. Perhaps the capacity for leaps of intuition and creativity is the same capacity underlying leaps of sheer intractable bone-headedness.
Johnnyb, here’s a simple way you can demonstrate the usefulness of CSI.
1. Calculate the CSI of Mt. Rushmore, S. Dakota.
2. Calculate the CSI of Devil’s Tower, Wyoming
3. Use the values to demonstrate one is designed and the other isn’t.
Show us what ID can do. 🙂
Adapa,
I understand certain people may not find this persuasive. Thats up to them. The argument I made is a positive argument based on a standard of observation and evaluation to make an inference. Neither personal incredulity or ignorance is involved.