{"id":772,"date":"2009-05-11T14:17:00","date_gmt":"2009-05-11T14:17:00","guid":{"rendered":"http:\/\/scientopia.org\/blogs\/goodmath\/2009\/05\/11\/dembski-responds\/"},"modified":"2009-05-11T14:17:00","modified_gmt":"2009-05-11T14:17:00","slug":"dembski-responds","status":"publish","type":"post","link":"http:\/\/www.goodmath.org\/blog\/2009\/05\/11\/dembski-responds\/","title":{"rendered":"Dembski Responds"},"content":{"rendered":"

Over at Uncommon Descent, Dembski has responded to my critique of
\nhis paper with Marks. In classic Dembski style, he ignores the
\nsubstance of my critique, and resorts to quote-mining.<\/p>\n

In my previous post<\/a>, I included a summary of my past critiques of
\nwhy search is a lousy model for evolution. It was a brief summary of
\npast comments, which did nothing but set the stage for my
\ncritique. But, typically, Dembski pretended that that was the entire
\nsubstance of my post, and ignored the rest of it. Very typical of
\nDembski – just misrepresent your opponents, create a strawman, and
\nthen pretend that you’ve addressed everything.<\/p>\n

<\/p>\n

Dembski does his best to misrepresent even that small portion. As
\nI’ve said lots of times in the past: search is a crappy model for
\nevolution. It’s value is that it provides a handle on which to hang
\nyour intuition. But its great weakness is that your intuition is
\nfrequently wrong<\/em>. Our sense of a landscape is something
\nstatic and unchanging, smooth, with hills and valleys. We intuitively
\nexpect a landscape to have hills with maxima, and valleys with minima.
\nThat comes both from our intuition about “normal” shapes, and our
\nexperience with 3 dimensional space. But in evolutionary search,
\nwe’re looking at a dynamic landscape with thousands of dimensions.
\nNone of our expectations work there.<\/p>\n

Dembski points out that you can always add time as a
\ndimension. That’s true – in fact, I’ve pointed that out before.<\/a> But
\nthe point still holds that the static landscape model doesn’t work
\nvery well. Why? The simple way of adding time as a dimension
\ncorresponds to our experience of time: a linear progression forwards –
\na single additional dimension, where the motion through that dimension
\nis linear. But if we want to model evolution, we can’t<\/em> use
\nthat model of time – because the landscape changes in response to
\nour search<\/em>. If the search chooses step A, then the landscape
\nresponds in one way; if the search chooses step B, the landscape
\nresponds in a different<\/em> way. <\/p>\n

To make that a bit more concrete, think about a simple example
\nfor which we’ve got lots of observations: antibiotic resistance.
\nYou’ve got a bacteria that’s not resistant to any antibiotics. How’s
\nit going to evolve over time? No particularly good idea. Now,
\nyou add penicillin to its environment. What’s going to happen? You’re
\ngoing to select for penicillin resistance. Now, imagine two scenarios:
\none where you keep giving penicillin, and one where you stop the
\npenicillin. In the first scenario, the landscape requires that the
\nbacteria maintain penicillin resistance; the bacteria can’t reproduce
\nif it isn’t resistant. So you’ve got a landscape that maximizes
\nthe survival of the resistant bacteria. In the second scenario,
\none possible outcome is that resistance disappears<\/em>: the
\nresistant bacteria waste energy on their resistance strategies,
\nand get outcompeted by the non-resistant. The two landscapes are
\ntotally different.<\/p>\n

In that case, you’ve got two different landscapes because of two
\ndifferent external interventions. But you can get an equivalent
\nsituation without the external change of removing the
\npenicillin. Bacteria have multiple ways of combatting antibiotics.
\nPenicillin works by interfering with the process by which the bacteria
\nproduce their cell walls. When they divide in the presence of
\npenicillin, they basically explode – they split, and they can’t
\nmanufacture the walls to close the two new cells. Some bacteria
\nrespond to this by producing a chemical that binds to the penicillin
\nmolecule, so that it’s neutralized, and can’t interfere with the
\nproduction of the new cell wall. Another response is to build the
\nwalls in a different way, so that penicillin no longer
\ninterferes. Both of those strategies work; and they each have
\nadvantages and disadvantages. When exposed to penicillin, bacteria
\ncan go either direction. If they manufacture their cell walls
\ndifferently, that produces one fitness landscape; if they excrete
\na penicillin neutralizer, that produces a different fitness
\nlandscape.<\/p>\n

The simple way of adding dimensions to a landscape model
\ndoesn’t capture this. And it’s really quite difficult to actually
\ncapture this in a landscape model – it basically means that
\nyou can never know what the landscape is like<\/em>. You get
\na massively multidimensional landscape, with absolutely no ability to
\npredict what will happen as time passes. It basically turns the model
\nwhose advantage is simplicity and intuitiveness, and turns it into
\nsomething astonishingly complex and non-intuitive.<\/p>\n

In short, it’s a crappy model.<\/p>\n

But – none of that matters for this discussion<\/em>. I don’t
\nlike reasoning about evolution using search, because I find it to be
\nobfuscatory rather than clarifying; but you can<\/em> model
\nevolution as a search over an incredibly complex and
\nhard-to-comprehend landscape. That’s why the critique of the landscape
\nmodel in my original post was three sentences out of a rather long
\npost.<\/p>\n

When you look at Dembski’s article, it’s very messy because he
\nchose (deliberately) to use a messy model. But the messiness isn’t the
\nproblem: the messiness is a distraction, which is intended to obscure
\nthe actual problem. And the actual problem is that it’s a circular
\nargument. It basically starts by assuming that only an intelligent
\nagent can create information; then it jumps through all sort of hoops
\nin order to finally conclude that only an intelligent agent can create
\ninformation. In that sense, the whole thing is completely content
\nfree: it’s pure obfuscatory math – math used to obscure an argument,
\nrather than clarify or formalize it. In fact, the fundamental argument
\nof the entire thing is non-mathematical: “intelligence” isn’t a part
\nof the math at all.<\/p>\n

So what’s all that impressive looking math about?<\/p>\n

Not a heck of a lot. There’s a reason that I keep going on about
\nhow it’s all just a mass of obfuscation.<\/p>\n

First of all, it pretends to present a conservation law. Second,
\nit pretends to prove that an intelligent source of information is
\nrequired. But in fact, the conservation law isn’t really a
\nconservation law, and the idea that an intelligent source of
\ninformation is required is not only not proved by the argument based
\non the conservation law, but is in fact inconsistent with<\/em>
\nthe conservation law.<\/p>\n

We’ll start with the conservation law. As I said in the previous
\npost, Dembski engages in a whole lot of obfuscatory mathematics to
\ncreate a supposed proof that in order for a search to perform better
\nthan a random walk, that search must, in essence, cheat: it must
\nencode information about the landscape that it traverses. In Dembski’s
\nterminology, it must contain “smuggled” active information. The
\nsupposed “law of conservation of information” basically says that the
\namount of information encoded into the search can be measured by the
\nthe rate at which that search outperforms random walk.<\/p>\n

So why is that not<\/em> a “conservation of information”
\nlaw?<\/p>\n

Because there’s no conserved quantity. In a real conservation law,
\nyou have a measured quantity that you start with, and throughout any
\nseries of actions or events, you can prove that that quantity never
\nchanges. For example, you can look at a physical system in a
\nparticular frame of reference and measure the total momentum in the
\nsystem. Then throughout any series of interactions, you can show
\nthat the momentum never changes.<\/p>\n

In Dembski’s system, can you measure the total information in the
\nsystem? No. Can you show that the amount of information in the system
\nis the same before and after a search? Not in any meaningful way,
\nno. Can you look at a search function, and ask how much information
\nit encodes from a particular landscape? Not in any meaningful way,
\nno. <\/p>\n

To be a little bit concrete: there is no<\/em> analytic way to
\nlook at a search function and quantify how much &lquot;active
\ninformation&rquot; is embedded in it. It can only be determined
\nretrospectively: run the search in the landscape, determine how well
\nit performed, and then quantify its performance. Looked at from that
\nperspective, it’s a (sloppy) re-statement of Kolmogorov-Chaitin information
\ntheory: the information contained in a string (or, to use Dembski’s
\nscenario, a landscape position) is the shortest program that can
\ngenerate that string (or a path to that position).<\/p>\n

So – Dembski unknowingly rephrased a bit of K-C theory. That’s not
\nso bad, right? To manage to redo a bit of work by two of the best
\nmathematicians of the 20th century? Well, if that’s what he meant to
\ndo, it wouldn’t be bad. But it’s very bad for Dembski’s argument:
\nK-C theory doesn’t support Dembski’s argument. In fact, in K-C theory,
\nyou can’t quantify<\/em> information in a precise way. Beyond
\nsome absolutely trivial examples, you can’t measure the quantity of
\ninformation.<\/p>\n

Dembski is arguing that information must be conserved, using a
\nframework that in which you can’t measure it<\/em>. And further,
\nit’s a framework in which the intuitive<\/em> notion of information
\n– which is what Dembski is really relying on – has absolutely no
\nconnection with the information that’s supposedly hidden in
\nthe search.<\/p>\n

Once again, I’ll get a bit concrete. A while back, I wrote about a
\nNASA
\nexperiment to design a better antenna.<\/a> The engineers involved used
\nan evolution-based approach. The result was completely
\nunexpected<\/em>. The particular shape that turned out to optimal was
\nnot something that any of the engineers involved would have come up
\nwith by themselves. By Dembski’s argument in this paper, the
\ndescription of that antenna was encoded into the search system as
\n“active information”. But the people involved in the experiment
\ndidn’t have<\/em> the information about what the optimal antenna
\nwould look like. And by looking at their simulation, no one could
\nextract or identify the information that Dembski insists was smuggled
\nin. Running the experiment produced the information that this
\nparticular antenna appears to be optimal. In the Kolmogorov-Chaitin
\nsense, that means that information about the optimality was contained
\nin the combination of the search and the search landscape – which is
\ntrivially true, because running the program in the landscape produces
\nthat antenna as a result. But that’s not<\/em> what Dembski is
\nclaiming to say. Dembski is saying that the program implicitly
\ncontains the solution.<\/p>\n

Why does it implicitly contain the solution? Because Dembski says
\nso. Seriously – that’s what his argument reduces to. He
\ndefines<\/em> the active information in a system in terms of how
\nthat system performs in a search. Then he shows that the amount of
\ninformation that results from doing the search is equal to the amount
\nof active information in the search algorithm. It’s a trick of
\ndefinitions, obscured by a lot of pointlessly complex math. In
\nessence, it reduces to making a blind assertion: information is
\nconserved; therefore any system that can in any sense produce
\ninformation must contain that information. But since by the
\nalgorithmic definition of information, any system that produces
\ninformation contains the information it produces, saying that
\ninformation is conserved is a simple tautology – exactly the kind of
\nstatement that Dembski mocks in the beginning of the paper!<\/p>\n

Moving on to the second point, Dembski pretends that this whole
\nargument somehow shows that intelligence must be involved in any
\nprocess that creates information. As I said in both the previous
\nparagraph and the previous post, it doesn’t work – because he doesn’t
\nactually make that argument. He assumes<\/em> it as a premise
\nbefore he starts, and then he concludes it at the end as if it’s a new
\nresult. Look at all of his mathematical arguments – there’s nothing in
\nthere that defines “intelligence”. There’s nothing that concludes
\nanything about intelligence. There’s the whole so-called conservation
\nlaw – but it doesn’t allow any exceptions<\/em>. There is nowhere
\nin that whole line of mathematical argument that says “Intelligence
\ncan create information”; in fact, according to that argument,
\nintelligent agents cannot<\/em> create information – nothing
\ncan.<\/p>\n

That’s the nail in the coffin of Dembski’s argument. His
\nwhole conservation law effectively argues that you can’t create
\ninformation. He claims<\/em> that it says you can’t create
\ninformation without intelligence – but that exception, that
\nintelligence can create information – is totally omitted from
\nthe math<\/em>.<\/p>\n

You can apply Dembski’s own argument to an intelligent
\nagent searching for a solution to something. The intelligent
\nagent, by Dembski’s own argument, already contains all of
\nthe information<\/em> that it uses in the search. If
\nyou actually follow through on what Dembski’s conservation of
\ninformation stuff says, what it ultimately means is that
\nintelligent agents can’t produce information<\/em>. If an
\nintelligent agent produces information, they must contain that
\ninformation as well. So either intelligent agents can’t create
\ninformation, or information isn’t conserved. Either way,
\nDembski winds up defeating his own argument.<\/p>\n

Dembski’s entire argument is self-defeating<\/em>. Either
\ninformation is conserved in the sense that he insists – and then
\nintelligent agents can’t produce information; or intelligent
\nagents can<\/em> create information – and then the conservation
\nof information “law” is refuted.<\/p>\n

As I said in the original post, and have now argued multiple times
\nhere, this is all an eloborate, highly obscured circle. Information is
\nconserved, which means that information can’t be created; therefore
\nsomething must have created it. Why? Because at the beginning of the
\nwhole argument, he asserted that an intelligent agent can create
\ninformation; therefore if you can find any information, since it’s
\nconserved, something intelligent must have created it. He defined
\ninformation as a conserved quantity created by an intelligent agent,
\nand then used all of that impressive-looking math to ultimately
\n“prove” that information is a conserved quantity created by an
\nintelligent agent. It’s a perfect example of obfuscatory math: all of
\nthat math is just a smoke screen to cover up the fact that he’s
\nembedded his conclusion in the assumptions at the beginning of the
\nargument – and worse, it’s an argument that contains its own
\nrefutation, because by the supposed conservation law, an intelligent
\nagent can’t create information either – so no information can ever
\nbe created.<\/p>\n

You can see the foundation of the basic circle of the argument if
\nyou look at the dreadful text of section one of the paper. He starts
\nwith Shannon’s definition of information, and then (as I pointed out
\nbefore), engages in a bunch of sillyness to try to pretend that
\nvarious philosophers who talked about philosophical ideas of
\ninformation were actually talking about Shannon information, and then
\nuses them to build up a purely philosophical argument that only
\nintelligence can create information. Then he uses that as a basis of
\nhis further arguments.<\/p>\n

The whole paper is an exercise in circularity. There’s nothing
\nthere – which is why this isn’t a paper in a mathematical journal;
\ninstead, it’s just a chapter in one of Bill Dembski’s vanity
\npublications.<\/p>\n","protected":false},"excerpt":{"rendered":"

Over at Uncommon Descent, Dembski has responded to my critique of his paper with Marks. In classic Dembski style, he ignores the substance of my critique, and resorts to quote-mining. In my previous post, I included a summary of my past critiques of why search is a lousy model for evolution. It was a brief […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"jetpack_post_was_ever_published":false,"_jetpack_newsletter_access":"","_jetpack_dont_email_post_to_subs":false,"_jetpack_newsletter_tier_id":0,"_jetpack_memberships_contains_paywalled_content":false,"_jetpack_memberships_contains_paid_content":false,"footnotes":"","jetpack_publicize_message":"","jetpack_publicize_feature_enabled":true,"jetpack_social_post_already_shared":false,"jetpack_social_options":{"image_generator_settings":{"template":"highway","default_image_id":0,"font":"","enabled":false},"version":2}},"categories":[16,31],"tags":[],"class_list":["post-772","post","type-post","status-publish","format-standard","hentry","category-debunking-creationism","category-intelligent-design"],"jetpack_publicize_connections":[],"jetpack_featured_media_url":"","jetpack_shortlink":"https:\/\/wp.me\/p4lzZS-cs","jetpack_sharing_enabled":true,"jetpack_likes_enabled":true,"_links":{"self":[{"href":"http:\/\/www.goodmath.org\/blog\/wp-json\/wp\/v2\/posts\/772","targetHints":{"allow":["GET"]}}],"collection":[{"href":"http:\/\/www.goodmath.org\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/www.goodmath.org\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/www.goodmath.org\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/www.goodmath.org\/blog\/wp-json\/wp\/v2\/comments?post=772"}],"version-history":[{"count":0,"href":"http:\/\/www.goodmath.org\/blog\/wp-json\/wp\/v2\/posts\/772\/revisions"}],"wp:attachment":[{"href":"http:\/\/www.goodmath.org\/blog\/wp-json\/wp\/v2\/media?parent=772"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/www.goodmath.org\/blog\/wp-json\/wp\/v2\/categories?post=772"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/www.goodmath.org\/blog\/wp-json\/wp\/v2\/tags?post=772"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}