----------------------
George Noonan wrote:
>>The recent issue of First Things has an interesting article by Willaim
>>Dembski, entitled Science and Design in which he proposed the mathematics of
>>SETI (Search for Extra-Terrestial Intelligence) to confirm Behe's irreducible
>>complexity as a criteria for biological design. Any comments?
Based on Dembski's earlier work, I expect that his new work provides
some carefully argued criteria for "detecting design" in biology. (By
"design," Dembski means extra-natural assembly. Dembski's use of the
term "design" excludes self-assembly of biological complexity. It even
excludes self-assembly via natural laws which were themselves designed
to produce self-assembly.) I expect that Dembski's proposed criteria
for detecting design are sound.
Whether or not irreducibly complex biochemical structures actually
*meet* those criteria is another question. To establish design (extra-
natural assembly) by these criteria, we must first calculate the
probability that these biochemical structures could have evolved. At
present, calculating those probabilities accurately is beyond our
capabilities. I don't know whether or not Dembski's new article (or the
new book he is co-authoring) claims to have *actually* detected design
by these criteria; if so, such claims are necessarily premature.
Dembski's earlier work on detecting design offered the twin criteria of
"small probability" and "pre-specification." His new work, apparently,
changes those to criteria to "complexity" and "specification."
("Complexity" is closely related to --- if not identical to ---
information content.) I'll expand on those two criteria below, using
Search for Extraterrestrial Intelligence (SETI) as an example. At the
end of this post, I'll return to biology.
Suppose you monitor a radio signal from a distant solar system. If the
signal is short (or if you monitor it for a short time), such that the
signal could be represented with just a few hundred bits of information,
the signal is not considered "complex." If the signal requires billions
of bits (or more) to represent, it would be fair to call it "complex."
Complexity alone is not enough to establish design. Any random signal,
monitored for a long time, is complex. But suppose that a complex,
random-appearing radio signal has been pre-specified. Suppose, for
example, that someone handed you a data tape the day *before* you
monitored the radio signal. (Suppose he told you that space aliens had
given him that tape.) Suppose you found that every one of the billions
of bits, on the tape and in the radio signal, were an exact match. In
that case, the complex radio signal had been pre-specified. You would
be justified in concluding that intelligence was behind the event. (In
this particular case, you would probably guess that *human* intelligence
was behind it; you would work very hard to figure out whether some human
source had transmitted the radio signal.)
Note that you need both pre-specification and complexity to establish
design. If you take the example above, but reduce the complexity of the
radio signal to a few bits of data, you would probably conclude it was a
case of "lucky guess" rather than "design."
Pre-specification is unlikely to happen in SETI, or in biology. Is it
possible to have "specification" without pre-specification? I believe
so. You have "specification" without pre-specification if you meet
three criteria: Pattern, Probability, and Pay-off. (Note: These are
my own terms, not Dembski's, although Dembski may use similar terms and
concepts in his discussion of specification.)
Pattern: SETI researchers are looking for a signal which contains a
recognizable pattern --- for example, a series of pulses which encode
the first 10000 prime numbers, or a signal which has the mathematical
characteristics of a symbolic language (rather than the mathematical
characteristics of a purely stochastic process).
Probability: If a pattern is detected, we must determine the
probability that a purely natural process could produce it. A complex
radio signal which included a small snippet of a recognizable pattern
(for example, the first ten prime numbers) might be considered a
candidate for SETI, but would not be considered convincing. An
*extremely* low-probability event, such as encoding the first ten
thousand prime numbers, is required to be convincing.
Pay-off: In general, intelligent agents don't go to the trouble of
producing lengthy, complex patterned signals unless doing so has some
benefit (pay-off) for themselves or for others. Conversely, if a
complex, patterned signal *does* provide some clear pay-off for
somebody, we suspect it is designed. In the case of SETI, there is an
obvious pay-off from inter-stellar radio communication.
To further illustrate the "pay-off" criteria, I'll adapt an example from
an earlier paper by Dembski. Suppose the government institutes a
lottery which operates in the following manner: A machine rapidly
performs 1000 quantum measurements to generate a truly random 1000-bit
"winning" lottery number. Simultaneously, this same machine uses
quantum measurements to generate random 1000-bit "ticket" numbers for
every citizen. The odds that anyone's ticket number will match the
winning number are so small as to be not worth considering. If there
was a winner, we would be justified in concluding that the system had
been tampered with. On the other hand, if there is no winner to the
lottery, we wouldn't bother investigating the possibility of tampering.
"Pay-off" is an indicator of "design."
Pattern, probability, and pay-off are not rigidly defined. Sometimes,
we can erroneously think we see patterns in randomly produced signals.
Sometimes, it is not clear whether or not a pattern could have a "pay-
off." Calculations of probability always have assumptions. We can only
calculate probabilities based upon the natural processes which we know.
If we do detect a pattern, could it be generated by an unknown natural
process?
Pattern, probability, and pay-off are not *all three* absolutely
required to conclude that something has been "specified." In the case
of the hypothetical government lottery, pay-off and probability are
enough to conclude design, even if the 1000-bit winning number has no
particular pattern. In other cases, pattern and probability may be
sufficient, even when pay-off is debatable, to conclude specification.
Taken together, however, a highly recognizable pattern, a clear pay-off,
and a low probability of being produced naturally provide a very strong
argument for "specification" and "design."
The discovery of pulsars provides an example of how pattern,
probability, and pay-off can interact. When the first pulsar was
discovered, there was no widely-accepted natural process that could
produce such a radio signal. (Given widely-accepted natural processes,
the pulsar signal was a low-probability event.) The hypothesis was
entertained that "little green men" produced this signal. The signal
was not very complex, and the possible "pay-off" for producing such a
simple signal was not considered to be high, so the LGM-hypothesis was
not taken very seriously. When a second pulsar was discovered, the LGM
hypothesis was dropped. The signals were simple, and what possible pay-
off could there be in producing two or more of them? Pulsars were
quickly taken to be evidence of a new natural phenomenon, neutron stars,
which until then had been considered highly theoretical.
Now back to biology. What about life and the DNA sequences which encode
it? Does it meet the criteria of complexity and specification? It's
complex. Is it specified? I believe it meets the criteria of pattern
and pay-off. The DNA sequences of living organisms are not random;
there is a pattern. The pay-off (to us, at least) is obvious. But can
we calculate the probability? Ay, there's the rub! Not all possible
DNA sequences are equally probable. How can we empirically calculate --
- with realistic assumptions --- the probability that they self-
assembled via natural mechanisms? At present, we can't do that
calculation with realistic assumptions.
One of the Dembski's goals is to argue that, in biology and
biochemistry, the design hypothesis (extra-natural assembly) should not
_a_priori_ be ruled out. I agree. Design should not be excluded based
on simplistic methodological claims of how science operates. Design
should not be excluded based on claims that "science can't detect
design." Science alone cannot detect design. Science can calculate
probabilities. Science can be used to detect patterns. Science cannot
detect "pay-off," but in the case of biological life that isn't even a
point of contention. As the example of SETI demonstrates, scientific
methods and arguments *can* be used to help build a strong case for
design.
I hope that Dembski is suitably cautious when offering probability
calculations for abiogenesis and the evolution of complexity. Overly-
simplistic and already-refuted calculations, which claim to demonstrate
small probabilities, are a dime a dozen. It would be a mistake to use
such calculations to claim that the design hypothesis has been *proven*.
Such premature claims would create unnecessary arguments for Dembski and
draw attention away from the goal of establishing design as a valid
hypothesis.
Loren Haarsma