Per Gregory and Dr. Felsenstein’s request, here is an off the top of my head listing of the major theories I can think of that are related to Dembski’s ID theory. There are many more connections I see to mainstream theory, but these are the most easily connected. I won’t provide links, at least in this draft, but the terms are easily googleable. I also may update this article as I think about it more.
First, fundamental elements of Dembski’s ID theory:
- We can distinguish intelligent design from chance and necessity with complex specified information (CSI).
- Chance and necessity cannot generate CSI due to the conservation of information (COI).
- Intelligent agency can generate CSI.
Things like CSI:
- Randomness deficiency
- Martin Löf test for randomness
- Shannon mutual information
- Algorithmic mutual information
Conservation of information theorems that apply to the previous list:
- Data processing inequality (chance)
- Chaitin’s incompleteness theorem (necessity)
- Levin’s law of independence conservation (both chance and necessity addressed)
Theories of things that can violate the previous COI theorems:
- Libertarian free will
- Halting oracles
- Teleological causation
In fairness, Dembski at least tried to calculate the CSI of the bacterial flagellum in No Free Lunch, which featured the flagellum on its cover.
He failed, rather spectacularly, and no one has succeeded in the meantime. I’ve mentioned Dembski’s failure a couple of times in this thread, but it appears Eric doesn’t want to talk about it. Doesn’t exactly fit with his narrative.
“In the inversion”? The inverse of the permutation function contains no information about the like-a-flower image. I could give it to you in detail, and you would never be able to figure out the image from it.
What, “goodbye” again?
Well, I will post a discussion of ASC, and why any conservation of it does not imply that Dembski’s 2002 CSI is conserved, at Panda’s Thumb sometime soon.
And I hope to follow that by going back to my 2012 argument at TSZ.
I hope you’ll pay enough attention to grapple with those too.
The problem is that there’s no legitimate reason for the “transfer”. Dembski stated his case without referring to ASC or Montañez or “normalization by kardis”. Joe also stated his counterexample without recourse to any of those. If Dembski’s argument is correct, then Eric should be able to defend it on its own terms.
Joe’s counterexample is simple and straightforward:
1. A flower photo contains CSI with respect to a “looks like a flower” specification.
2. Apply a permutation function f, and the resulting image no longer contains CSI with respect to that same specification.
3. Apply the inverse of f, and the CSI magically reappears, despite not being contained in f or the inverse of f.
4. Therefore, CSI is not conserved.
If Eric is right, then there must be an error in Joe’s reasoning. Which step is in error, and what precisely is that error?
If Eric could find one, he wouldn’t be so eager to flounce.
The image was “carefully constructed” as follows: I took a photo of a Zinnia flower on our patio. I popped that into a photo editing program, reduced it to 101 x 100 pixels, and thresholded those to make a black and white image of 10,100 bits. I could have done the same with any other photo of a flower. The objective of this process was to make an image that was about 100×100, and when thresholded to black-and-white pixels satisfied the specification “like a flower”. Other than that it was not very carefully constructed.
Eric, to Joe:
Also, what’s this about “high CSI segments”? Eric seems quite confused.
Joe’s permutation (and its inverse) could be applied to a photo of a bicycle, or a tree, or Barack Obama, with the same results: a decrease in CSI when the permutation is applied, and a restoration of CSI when the inverse is applied. The CSI is not in the permutation function or its inverse.
CSI is not conserved.
I choose not to talk about it because I don’t know much about the details of biology. I choose to talk about areas in my expertise.
You don’t need to be an expert in biology in order to understand Dembski’s argument or to see why it fails. Dembski himself is no biologist.
As someone who is quite gullible when it comes to ID (and other) claims, you could benefit from paying more attention to ID’s failures, including Dembski’s.
The function contains exactly the information necessary to map the specific scrambled image back to the flower. That is the CSI.
Nope, it’s not the CSI. You’ve changed the specification, just like Dembski did. That’s the mistake Joe has been explaining to you:
That’s funny. That same reverse permutation also contains the CSI for a picture of Abraham Lincoln, a picture of a bunny rabbit, the flag of British Columbia, and a lot of others.
It’s a sign of the transcendent unity of all things. Praise Jesus.
Thank you for reminding me that the concepts did not originate with the ID people. I am not familiar with the original work you mention, but am I right in thinking that those authors never concluded that CSI or Functional Information are indicators of ‘Design’?
As you say, ID has been arguing furiously for 20 years now against a position that nobody holds: that evolution is random. Evolution clearly isn’t random, it is precisely the remarkable adaptation of organisms to their environment that needs to be explained. Nothing random about that. Surely evolution is guided – guided by the environment. In a truly random universe, evolution would be impossible (‘random’ in the sense of unorganised and devoid of predictable processes).
Whereas, if your original picture was just a matrix of random pixels without resemblance to anything we humans would recognise, that very same reverse permutation function would contain no CSI at all.
I would really like to see Eric or someone from the ID side calculate the CSI of your permutation function!
But they won’t, won’t they? They never do.
“The backwards reasoning from a mathematical model to reality is inherently perilous, because mathematics can model all sorts of unphysical and counterfactual things.””
I’m only trying to understand the logic of Eric’s mathematical argument. The Kraft-McMillan inequality he brings up does result in a probability distribution which I assume he is somehow linking to the probability distribution involving f that he explains in his latest reply to me. So I think that is a least partly why he thinks can turn your argument into one about KC (as used in ASC) and then refute your argument there. But that application of the inequality is where he loses me.
But I should emphasize that this is just an intellectual exercise for me. Whether math models have any scientific merit is a different question. In particular, I think the scientific, biological issues for Dembski’s various pieces of work have been thrashed out many times at TSZ — whether evolution is like a search; the physical nature of the search space if it is considered a search; and of course the biology needed to compute the chance hypothesis in “event’s probability according to a chance hypothesis: P(X|C)”. No doubt, there are others that you know better than me.
I’ll look forward to your post on the scientific merits of ASC. If you are interested, there was an exchange between Joshua and Eric on PS about the mathematical merits of ASC
Related discussion here
ETA Clarity in “chance hypothesis” for P(X|C)
See my reply to Joe.
ETA: Specifically, I read Eric as thinking there is a legitimate reason for the transfer, and that reason is part of his claim to have disproven Joe’s argument. I am trying to understand why he thinks that.
Thanks for the links.
<sarcasm> Yeah, tell me about it </sarcasm>. George Box famously put it this way: all models are wrong; some models are useful.
Seeing whether a proposed generality works in a special case — in a mathematical model, for example — is useful, for there we can actually evaluate it. In general too little theory is done in biology, and the deficiency of theory is increasing. I see that in the difficulty of getting biological sciences departments to teach any courses about theory or to require students to take any from other departments.
Note that when Eric finally got around to answering Joe’s argument on its (and Dembski’s) own terms, he failed:
Okay. Here’s the paper: http://bio-complexity.org/ojs/index.php/main/article/view/BIO-C.2018.3/BIO-C.2018.3
Bio-complexity though :/ who were the peer reviewers? Would Tom or Joe like to offer their thoughts? What are the entailments of this? Will the designer be coding new modules soon? What analogs do we have with human coding endeavors?
You ask: “Additionally, what has Darwinism/Materialism provided that is of value?”
All of science.
“Everything valuable in modern biology depends on computational and information theory constructs, which are the result of an ID consistent paradigm, not a D/M paradigm.” Well if you mean math, then yes. I thin your overselling this here fine royal garment. Can you give something *specific*?
But isn’t the second specification related to the first through the specification of f, in the KC sense of specification, which is how Eric describes it? Eric does mention the CSI(f), which includes f’s probability (or the image’s?) and f’s specification, so I think that is a hint worth pursuing.
But that is enough work for my senior citizen brain for today. Time for me to catch up on last night’s HBO (yes, I still record stuff that occurs past my bedtime; I know about streaming but I avoid that new-fangled stuff if I can).
The two specifications are related but distinct, and that distinction makes all the difference in the world, since SI/CSI depends intimately on the specification. The specification needs to be held constant, and Eric is not doing so.
Also, it’s absurd for Eric to argue that the reverse permutation function contains the CSI. The permutation and its inverse are generated randomly and contain no information whatsoever about the content of the image.
If he disagrees, he’s welcome to demonstrate how one can derive the zinnia image from the reverse permutation function alone.
It can’t be done.
Eric may be getting the KC math wrong — of the TSZ contributors, Tom seems best qualified to judge that. But I do think Eric can follow your/Joe’s arguments. Yet he claims the arguments are mistaken.
Based solely on his post on this thread, that is ignoring Demski’s papers, I’d like to understand why he thinks that, and maybe learn a bit more about KC in the process.
Yes, the specification is different for demonstrating that f contains CSI.
Using ASC notation, and assuming f is essentially the XOR mask of say 100×100 bits, and X is the scrambled b/w image of 100×100 pixels, then I am adding X to the specification:
ASC(f, P, C) = -log2 P(f) – K(f|X,C)
As you point out, if we kept the specification constant, then there is zero ASC in f.
Since we are discussing what function can convert X back into the flower photo, it makes sense to add X to the specification.
No, the function f is a permutation of the numbers 1 to 10,100 (the size of the image was 101×100). A permutation is not an XOR mask, it is given by a list of the numbers 1 through 10,100, each integer in that range being given once.
(Thinking of it as part of the specification of the image of a flower is bizarre, since the image could also be any other image, such as a photo of any other flower besides that zinnia).
Or, the original could be indeed an image of nothing at all, just abstract shapes or even random noise. In which case it makes no sense whatsoever to assign CSI to the function or to the original image.
So, it all depends on the observer and if she recognises something in the image. How can you capture that in formulas? As an objective measurement this approach seems useless.
If we are trying to get an image that is “like a flower” then it will mostly be images of flowers that will work.
The more relevant issue for evolutionary biology is a specification that is fitness. I hope to get to that, using the calculations in my old 2012 TSZ post on that as an example instead of flowers.
All of this seems like arguing from the map what features exist in the territory.
Show me the chemistry that can’t happen. Show me the place that requires the magic wand.
I give Behe credit for trying. In the universe of ID, he is the only one who actually looked for a difficulty.
You’ll pardon me for concentrating on equations for gene frequencies and such. Us theoreticians get continually told that we are being unreal and should stop doing theory and concentrate on the details of real biology, or real chemistry. We get allergic to such statements.
Dembski’s argument was a theoretical one supposedly showing that one could not get genomes of high fitness unless Design intervened. Equations about gene frequencies, genotype frequencies and fitnesses are enough to refute it without getting all chemical.
… and more to the point, ID advocates think that they have shown by a theoretical argument that there must be some-kind-of Design Intervention. When you call for the details, it looks as if you are quietly admitting that their theoretical argument works. But it doesn’t work.
Old lawyer joke:
When you have the law on your side, pound the law.
When you have the facts on your side, pound the facts.
When the law and the facts are against you, pound the table.
Substitute chemistry, statistics.
Not really. The fact that they can’t even provide one actual example when asked for is prima facie evidence that their theory is bunk.
The Newton of Information Theory? Compare how often Newton’s equations are actually used in practice (i.e. all the time) to how often ID equations are used (i.e. never). If it wasn’t so sad it would be funny.
You’re overestimating him. He just demonstrated that he doesn’t understand Joe’s simple argument. Joe specified that f is a permutation. Eric took it to be an XOR mask.
Worse still, Eric’s counterargument doesn’t work even if you make f a randomly generated XOR mask.
Such an XOR mask, like the permutation function, would contain no information about the image to which it was applied. The CSI cannot come from the XOR mask.
In my mind, a working definition of a random series would be one that has no information.
Just doing a bit of free association, it would seem that DNA , even without knowing what it does, has a great deal of information about where it came from.
Which is a huge problem for you, because conservation of CSI only makes sense if the specification remains unchanged. Your error, like Dembski’s, is to change the specification mid-stream.
Then the original zinnia photo will have no CSI according to that new specification.
CSI is not conserved.