Mar 31, 2011

In Which I Post a Comment on Uncommon Descent

There is a recent post up at Uncommon Descent talking about CSI: if effect, they are trying to understand what it is.  Interestingly enough, about halfway through the comments, we find this:
CSI is unique among the arguments of ID proponents in that it leads to positive, potentially testable claims. Every other ID argument I’ve seen is an attack on modern evolutionary theory, not explicit support for ID. Further, the claim that ID is a reliable indicator of intelligent agency, if it could be demonstrated, would be world changing. Based on this, I would expect CSI to be the most active area of research for ID proponents, with new calculations being published frequently. Indeed, this is what I was hoping to find when I first became interested in the topic from lurking here and on other blogs.
My preliminary conclusions from this discussion differ greatly from my initial expectations. It appears to me that there are at least four major problems with CSI as used by ID proponents here:
1) There is no agreed definition of CSI. I have asked from the original post onward for a rigorous mathematical definition of CSI and have yet to see one. Worse, the comments here show that a number of ID proponents have definitions that are not consistent with each other or with Dembski’s published work.
2) There is no agreement on the usefulness of CSI. This may be related to the lack of an agreed definition, but several variants, that are incompatible with Dembski’s description, and alternative metrics have been proposed in this thread alone.
3) There are no calculations of CSI that provide enough detail to allow it be objectively calculated for other systems. The only example of a calculation for a biological system is Dembski’s estimate for a bacterial flagellum, but no one has managed to apply the same technique to other systems.
4) There is no proof that CSI is a reliable indicator of intelligent agency. This is not surprising, given the lack of a rigorous mathematical definition and examples of how to calculate it, but it does mean that the claims of many ID proponents are unfounded.
When I took advantage of Denyse O’Leary’s kind offer to make a guest post, I fully expected a lot of tangential conversation in the comments. What I did not expect was for us to be nearly 200 comments in without anyone directly addressing the five straightforward questions I asked, without anyone providing a rigorous mathematical definition of CSI, and without anyone demonstrating how to calculate CSI for the scenarios I described.
I on the other hand did expect exactly that to happen, or would have had someone told me about it beforehand.  I've actually written, and have for a long time intended to post, an extensive series of musings on the ID movement.  The most interesting part for the purposes of this discussion was that I actually proposed a series of experiments that the ID proponents should be engaging in if they actually had a case that they were legitimately trying to prove.  Reasonable experiments, experiments whose results would definitely get my attention.  The idea that they are doing publicity and blogging rather than actually performing such experiments speaks volumes.

So I decided to participate in the discussion.  The last comment asked for an example of something that CSI might be able to do.  Well, that was as good an opener as you could hope for.  Here is what I wrote:
here’s your example:
Casey Luskin has rather strongly implied that random strings of letters don’t have CSI, whereas sentences do. So here’s a very simple test for CSI. I here present two strings in a simple letter-to-number substitution code. One of them spells a coherent sentence composed by myself, and one is a string of randomly generated letters. Run the CSI calculations, SHOW ME the CSI calculations, and based on that determine which is which:
Remember: it is very important that this is done on the basis of ACTUAL CSI CALCULATIONS and not ‘cheating’ of some sort. Thus, the number-substitution code. We’re testing your claim that you can detect design through calculating complex specified information, not your ability to paste text into Google translate, run expectation algorithms to determine real language consonant and vowel usage, look for repeating patterns, or even laboriously transliterate everything into the latin alphabet and read it aloud to see if it sounds like language.
The comment is still awaiting moderation; we'll see how this goes.

I rather doubt that anyone over at Uncommon Descent reads my blog, so I don't feel too worried about giving away the fact that the real sentence is the second one.  It says this: 
And right there is the problem with all their false analogies to language and meaning as 'information'.  Their knowledge that real sentences contain CSI is based solely on their knowledge of the Latin alphabet and the English language.  Take away that understanding and force them to base their prediction on actual mathematical calculation to differentiate between 'noise' and the letters of an alphabet that they don't know forming words that they don't understand, and they are utterly helpless.  Yet, this is precisely what they claim that their vaunted CSI ought to be able to do: distinguish between 'noise' and 'functional information'.  Well, time to put your money where your mouth is.  One of these strings is 'noise' and one is 'functional information.'  Get crackin'.


  1. Good for you for designing the experiment.

    Isn't CSI just an advanced version of Paley's "watchmaker" argument? If you define 'information' broadly enough, can't you find information anywhere - or reconfigure the information you have to get the information you want (e.g. The Bible Code books)?

    Have these people actually done work in semiotics enough to have an understanding of 'information'? Have these people approached loop quantum gravity and the holographic principle to understand how physics regards 'information'?

  2. ty, i've actually designed several. i might blog them at some point.

    yes it's sort of an obfuscated watchmaker argument, but it's more than that. it's the claim that there is a detectable property to data strings which is ONLY possessed by strings which were produced by an intelligence. to which I say, 'lol'. since Luskin has specifically talked about sentences and CSI, then if WERE true then you ought to be able to tell the difference between a string of random numbers and a string of numbers that correspond to letters that spell a sentence (sort of like ASCII but simpler). so, benefit of the doubt--if they actually have some calculation which can show this, then by all means let them demonstrate.

    'information' means whatever happens to prove their point at the moment. it's some sort of fudgy mushy thing that is TOTES DIFFERENT than Kolmogorov complexity or Shannon information and rather more like 'meaning' in a linguistic sense (in English, no less) yes still MATHEMATICALLY DEFINABLE.

    except, when you ask them to define it mathematically, as the guest blogger over at uncommon descent has done, they can't.