Re: [asa] Bayesian inference and design inference

From: David Campbell <>
Date: Mon Sep 21 2009 - 13:17:29 EDT

It may be of interest to note that Bayesian methods have become rather
popular in evolutionary analyses over the past decade or so. To be
blunt, this largely reflects the existence of readily available,
reasonably user-friendly programs to do it, but there are some good
theoretical grounds, and it can be done to big data sets within a
reasonable amount of time.

> Now take an example that has been cited by Dembski as "design
> detection".  It relates to the case (if I recall the details
> correctly) of Nicholas Caputo, who was in charge of arranging the
> ballot tickets in elections in one particular ward (state? - I don't
> know the correct term).  It is well-known that the name that appears
> first on the list of a ballot paper has an unfair advantage because
> lots of people are too stupid to check and just put an X on the first
> name on the list, irrespective of the party.  Hence lots would be
> drawn to determine which party was at the top of the list each time.
> It was found that under Caputo's direction, a democrat had appeared 40
> times out of 40.  He was convicted on the grounds that this was so
> unlikely to happen by chance, that he must have rigged it.  This is
> cited as a clear instance of design detection.

For what it's worth, the secondary source mentioning the example that
I've been reading said it was 40 out of 41 times. An unstated, but
critical, prior, to this assumption of design is presuming that an
equal number of Democrats and Republicans were available to choose
from. In local elections, often there is only one party that bothers
to run for a particular office. Another assumption is that there are
no confounding factors affecting the listing order. For example,
politics seems to run in families a lot. If names are placed
alphabetically on a ballot, and most of the Democratic candidates are
named Jones and most of the Republicans are named Smith, then an
inadvertent bias could be created. (Of course, the ballot producer
ought to know better, but without the prior knowledge that he does
know that people often just check the first box, we can't tell.)

In a more complex example, Tony Hillerman (known primarily for mystery
novels with a Navajo setting) provided an entertaining record of a
case in New Mexico where complaints about not receiving promised
payment for running in the election revealed the fact that, by
carefully working out how many votes were needed to win, a politician
succeeded in splitting more thoughtless votes away from his opponents
than from himself (e.g., the numerous voters who evidently randomly
selected among either the Hispanic names or the non-Hispanic names
both received additional names to choose from).

The case of design inference suffers from the lack of a comparable
clearly calculated set of priors. We can check the number of people
running and their affiliation in a given election, but we do not know
the number of possible ways to make a functional biological system nor
what the odds are on any of those ways happening. Additionally, are
the existence of the universe and various natural laws assumed as
priors or not?

> Now take the case of the intelligent design inference.  In the
> publicity for his new book, Stephen C. Meyer states that DNA is like a
> computer code with immense amounts of information.  In every case we
> know about, information implies an information giver, and a program
> requires a programmer.  Hence design (the nature of the designer
> remaining unknown and unspecified) is the best explanation.

At the ASA meeting, this seemed to be accompanied with a definition of
information that seemed designed to identify complex biological
systems as designed and containing information. Two particular

Only those biological events perceived as significant innovations
qualify as involving new information, so all the intermediate steps
that lead up to it are dismissed as not really new.

There is a confounding of information and information detection or
evaluation. For example, "the past week has been very rainy here" is
information that was generated at a physical level [i.e., disregarding
issues of general providence] by a weather system, not an intelligent
information giver or programmer. Detecting and responding to that
information may or may not involve intelligent action. For example,
creeks respond to this by rising; I respond by carrying my umbrella;
roofs respond either by leaking or by keeping the water out; etc. To
take an ID argument, it was asserted that a computer program cannot
generate more information than it contains. This is incorrect. A
computer program cannot recognize more information than it contains,
because it must include the recognition data within it. But it can
generate an unlimited amount of information that may be recognized by
something else. For example, i=0 i:=i+1 print i goto 2 will generate
as much of the counting numbers as you let it. A program to spew out
random letter sequences (or perhaps one with some basic rules like the
general importance of vowels) will generate recognizable words in
among all the gibberish. Moreover, the recognizer need not be that
complex, either-it can merely recognize something that works by its

In living organisms, we have a system-the copying of complex organic
molecules-that generates a practically unlimited range of possible
sequences. If this sequence produces a result that is compatible with
the constraints imposed by the environment and is able to produce
copies, then we have survival and reproduction. Thus, we have the
environment serving as an unintelligent detector of "meaningful",
i.e., functional, information and the imperfections of copying serving
as a means to generate novel information, regardless of functionality.

> But by definition, you don't KNOW about the existence of an unspecified designer - the fact that you don't say anything about the identity of the designer undermines the whole argument. <

Of course, if you select a particular known designer, then you can
factor that in. But a militant atheist who assigns a prior
probability of zero to the existence of a divine designer will
therefore not detect a designer no matter how improbable the object in
question is. TEs are generally dubious that the designer spent a lot
of time stepping in and helping the evolutionary process along with
miracles, whereas ID advocates often seem certain that's what a
designer ought to do.

But the usual ID claim is that someone who is totally agnostic will
take a look at their data and say "Wow! God did it!" That disregards
the issue of priors on the designer; it probably also has something to
do with the much greater credibility given to ID on average by people
with prior commitments to the existence of a designer than by those
who don't.

This also relates to the problem of the analogy of archaeology or SETI
to ID. We have a good idea of what early humans are likely and
unlikely to do, and we have a good idea of what non-human causes could
do. So we can look at a stone and determine whether it is likely to
be a primitive tool or an artifact of weathering or natural breakage.
If space aliens are generally similar to humans, then we have some
idea of what sorts of signals they are likely to make; we also know
what a wide range of unintelligent astronomical objects can do. A
pulsar can generate a regular series of radiation bursts, so their
initial designation as LGMs for Little Green Men has been abandoned.

Dr. David Campbell
425 Scientific Collections
University of Alabama
"I think of my happy condition, surrounded by acres of clams"
To unsubscribe, send a message to with
"unsubscribe asa" (no quotes) as the body of the message.
Received on Mon Sep 21 13:18:16 2009

This archive was generated by hypermail 2.1.8 : Mon Sep 21 2009 - 13:18:17 EDT