|
Author
|
Topic: The Significance of Dembski's Pulsar
|
Jules
Member
Member # 181
|
posted 11. June 2006 21:08
In an essay ("On the Very Possiblity of Intelligent Design") printed in the book The Creation Hypothesis, (1992 or so), William Dembski offerred a thought experiment, which I will try to summarize:
Imagine a pulsar a thousand light-years from here that is sending out morse code signals. And when we send it questions in morse code, we get immediate answers from it, very specific to our questions.
When I first read this essay, this is where I stopped reading, and figured that Dembski was using this to try to show that we could prove that an intelligent designer was supernatural. If I had read further, I would have realized I was wrong. Dembski willingly admits that this part of the scenario does not prove that the designer is supernatural. (It might only prove, for example, that light can travel faster than we thought it could.)
But what it does prove is something just as significant: We know that the messages from the pulsar are intelligently designed. And we know this, even though we would have no inkling as to how they were designed or who designed them.
At first, this might not seem very significant. But when one realizes that the most oft repeated criticism of ID is that it cannot tell us how the object or phenomenon was designed or who designed it, I think we can begin to grasp the importance of this hypothetical pulsar.
Now I may be wrong, but I don't think anyone would question that the messages from the pulsar are indeed intelligently designed. Why? I suggest it's because of their specified complexity. Had the electromagnetic pulses occasionally sent something that could have been interpreted as morse code for a word or two, I don't think we would have been impressed. And a long string of pulses, with no apparent pattern probably wouldn't impress us, either. But a long string of pulses, all of which can be interpreted as morse code would be evidence that would overwhelm the strongest skeptic, I think. And as far as I can tell, it is because of the extent of the messages specified complexity.
Now critics may object that there is a world of difference between the morse code of the pulsar and anything ID'ists point to in living organisms. Perhaps they are right. Perhaps not. But I think Dembski's pulsar removes the teeth from that old objection they like to throw in our faces.
IP: Logged
|
|
2ndclass
Member
Member # 1979
|
posted 12. June 2006 13:08
I agree that the ability to answer specific questions would indicate intelligence. Indeed, the ability to pass the Turing test could be used as an operational definition for intelligence.
I disagree that this conclusion is soundly based on specified complexity. I'll run through a pulsar example to explain why.
Imagine that it's 1967 and the first regular periodic radio signal from space has just been discovered. Its pulse width is 0.04 seconds, so we can express the signal in binary at a rate of 25 bits per second. We observe the signal for an hour, gathering 90000 bits of information.
To determine the amount of specified complexity in this information, we use Dembski's definition given here: SC = -log2( 10^120 * SpecRes(T) * P(T|H) ), where T is the pattern, H is the chance hypothesis, and SpecRes is the number of patterns that are as simple and as improbable as T. Nobody knows anything about pulsars yet, so our chance hypothesis is random noise, giving us a P(T|H) of 5*10^-27092. This probability isn't high. Nor is it intermediate. To be generous with SpecRes, we assume that all signals that repeat every 34 bits are equally simple. (This particular signal repeats every 1.337 seconds, which is about 33.4 bits.) This gives us a SpecRes of 1.7*10^10.
Putting it all together we get -log2(10^120 * 1.7*10^10 * 5*10^-27092) = about 90000 bits of specified complexity. Since this value is far greater than 1, we have a very solid design inference. False positive.
Given that SpecRes is inversely correlated with simplicity, a simple periodic signal generates more specified complexity than a signal that answers our questions in morse code, all else being equal. Since we've found specified complexity to be unreliable for the former signal, by what logic should we consider it reliable for the latter signal?
[Edit: I miscalculated P(T|H) by a factor of 25. The final answer is still about 90000 bits.]
[Edit 2: I added the two redundant sentences in bold above to appease someone who complained that I was skipping the first two nodes of the EF.]
[ 05. February 2007, 16:35: Message edited by: 2ndclass ]
IP: Logged
|
|
2ndclass
Member
Member # 1979
|
posted 12. June 2006 14:13
BTW, Jules, I agree with your main point. Gaps in the causal story do not render ID false or useless. (The same is true for evolutionary theory.)
IP: Logged
|
|
Jules
Member
Member # 181
|
posted 13. June 2006 20:34
Darn, hit the wrong button and deleted my whole reply. I'll take another crack at it tomorrow.
IP: Logged
|
|
Jules
Member
Member # 181
|
posted 15. June 2006 10:54
Well, maybe the day after tomorrow, which is today. So let me try again. Sorry, 2nd Class, but I'm mathematically challenged. But when we get a regularly repeating event -- in this case the regularly repeating electromagnetic pulses -- then randomness is ruled out by common sense. We look for either law-like or design-like explanations. If I recall, it was even suspected at first that pulsars were signals from extra-extraterrestrials. But since the signal was so simple and constantly repeated, a law-like explanation was sought (and I believed found). If an extra-terrestrial was trying to signal its existence, we would expect it to send something more complex. This was the reasoning used by Carl Sagan in Cosmos. Which is why he thought mathematics would be used, since it is probably the universal common language.
So you ask:
quote:
Given that SpecRes is inversely correlated with simplicity, a simple periodic signal generates more specified complexity than a signal that answers our questions in morse code, all else being equal. Since we've found specified complexity to be unreliable for the former signal, by what logic should we consider it reliable for the latter signal?
The answer in the case of Dembski's pulsar, is that we have a very recognizable pattern (morse code, matched to a human language), that is far too complex to be produced by chance or necessity. We would know this, even if the pulsar didn't answer our questions immediately (thus avoiding Turing's test). If it only sent out long, complex, meaningful messages in morse code, we would know an intelligent source designed it. And we would know this without having the slightest idea how it was accomplished.
I'm glad you agree that gaps in the causal story do not render ID useless. Too bad most ID critics think it does.
IP: Logged
|
|
Zachriel
Member
Member # 1793
|
posted 15. June 2006 11:37
Jules: "If I recall, it was even suspected at first that pulsars were signals from extra-extraterrestrials. But since the signal was so simple and constantly repeated, a law-like explanation was sought (and I believed found)."
The suspicion was very short-lived and no one would have taken it seriously without significant investigation. The signal not only repeated, but was slowing. This indicated a rotating object losing its rotational momentum, now known as a pulsar.
SETI, on the other hand, looks for a narrow-band transmission with a the distinctive doppler shift associated with a planetary orbit. Even if found, it would only be considered a candidate ETI.
Jules: "The answer in the case of Dembski's pulsar, is that we have a very recognizable pattern (morse code, matched to a human language) ... we would know an intelligent source designed it."
Yes, incredible as it may seem, if someone sends us a message in English, we would tend to think that it was made by someone with knowledge of English.
But if someone sends us the Fibonacci sequence, we might look for other explanations.
IP: Logged
|
|
2ndclass
Member
Member # 1979
|
posted 15. June 2006 14:07
Jules: quote:
But since the signal was so simple and constantly repeated, a law-like explanation was sought (and I believed found).
Yes. Had they applied Dembski's specified complexity criterion, they would have concluded that the signal was designed. Since Dembski's method is supposedly 100% free of false positives, they would have had no reason to seek a natural explanation. Fortunately, they didn't take the Dembski route. quote:
The answer in the case of Dembski's pulsar, is that we have a very recognizable pattern (morse code, matched to a human language), that is far too complex to be produced by chance or necessity.
I certainly agree. The problem is that Dembski's use of the word "complex" doesn't match the common sense notion of the word. Most people would say that a long string of zeros isn't complex, but for Dembski it would be considered complex if it's improbable according to our chance hypothesis.
Dembski has characterized specified complexity as "simplicity of description combined with complexity in the sense of improbability of the outcome.". So given equally improbable phenomena, the simplest phenomenon has the most specified complexity.
quote:
I'm glad you agree that gaps in the causal story do not render ID useless. Too bad most ID critics think it does.
Yes, I think your point is a good one. We should welcome whatever insights ID can give us into the origins of phenomena, and not complain that it can't tell us more. Whether those insights are reliable and meaningful is a separate question.
[ 15. June 2006, 14:08: Message edited by: 2ndclass ]
IP: Logged
|
|
Jules
Member
Member # 181
|
posted 16. June 2006 16:27
Zach writes:
quote:
Yes, incredible as it may seem, if someone sends us a message in English, we would tend to think that it was made by someone with knowledge of English.
But if someone sends us the Fibonacci sequence, we might look for other explanations.
Well, actually it's an aperiodic series of long and short electromagnetic pulses, that we notice can be matched to morse code for English (or some human language). And we would tend to think it was made by someone with knowledge of morse code and English, because of the improbability of this happening by chance or necessity. Which is what specified complexity is all about.
Yes, we might look for other explanations for the Fibonacci sequence (or prime number sequence) because we might not think that it has reached such a level of improbability to rule out chance and necessity.
2nd Class writes:
quote:
es. Had they applied Dembski's specified complexity criterion, they would have concluded that the signal was designed. Since Dembski's method is supposedly 100% free of false positives, they would have had no reason to seek a natural explanation. Fortunately, they didn't take the Dembski route.
But Dembski's filter includes two categories before we get to Design: Chance and Necessity. The repeating Pulsar signal only rules out chance, not necessity. Dembski would not have concluded "Designed."
quote:
Yes, I think your point is a good one. We should welcome whatever insights ID can give us into the origins of phenomena, and not complain that it can't tell us more. Whether those insights are reliable and meaningful is a separate question.
I agree.
IP: Logged
|
|
2ndclass
Member
Member # 1979
|
posted 17. June 2006 13:48
Jules: quote:
But Dembski's filter includes two categories before we get to Design: Chance and Necessity. The repeating Pulsar signal only rules out chance, not necessity. Dembski would not have concluded "Designed."
According to Dembski, necessity can be concluded only if there is a known natural phenomenon to account for the event. Since pulsars were unknown in 1967, necessity would be ruled out.
IP: Logged
|
|
Jules
Member
Member # 181
|
posted 17. June 2006 15:59
2nd Class:
quote:
According to Dembski, necessity can be concluded only if there is a known natural phenomenon to account for the event. Since pulsars were unknown in 1967, necessity would be ruled out.
You might have a good point, 2nd. I'm not sure what Dembski would have done in this case. Again, I think common sense would tell us to look for a law-like explanation. So perhaps Dembski's filter needs work.
I've been thinking about Zach's comments. I think they might give us some insights into the psychology of inferring design.
First, in the case of Dembski's pulsar, I don't think it is only the extreme improbability that makes us infer intelligent design. It's also that the phenomena in question is so closely associated with something we know to be associated with intelligence -- human language. It seems to be a combination of the two: extreme improbability and association with something we identify with intelligence. And together they overcome any sense of implausibility we might have about a pulsar sending an intelligently designed signal.
The second example, that Zach suggests -- the pulsar that has a Fibonacci sequence -- does not make us automatically infer design. Why? Is it just that it isn't as improbable? I don't think so. I think it's also because we do not necessarily associate the Fibonacci sequence with intelligence. We find it in other natural phenomena. And we still must overcome the immense implausibility of a signal from a pulsar being intelligently designed.
To summarize:
1) Dembsi's pulsar = extreme improbability + tight association with intelligence, which is greater than the implausibility of an intelligently designed pulsar signal.
2) Fibonacci pulsar = exteme improbability (though not as extreme as Dembski's pulsar), but no tight association with intelligence, which is not greater than the implausibility of an intelligently designed pulsar signal.
Now let's imagine that we discover there is a plausible way to control pulsars, so that one can make them send out whatever kind of signal they choose. Now the implausibilty of an intelligently designed pulsar signal has been greatly reduced. So that we get:
2a) Fibonacci pulsar = extreme improbability, but no tight association with intelligence, which is now greater than the implausibilty of an intelligently designed pulsar signal.
We would probably now regard the Fibonacci pulsar signal as being intelligently designed.
We can probably generalize from this example to the issue of ID. Specified complexity, by itself, may not be enough to convince us that intelligent design has taken place. Questions of association with intelligence and plausibility also come into play. And since there is a great deal of subjectivity involved in determining the strength of these other factors, it would explain why some people think the case for ID is very strong, while others think it is very weak.
IP: Logged
|
|
GSchultz
Member
Member # 1953
|
posted 30. July 2006 00:31
Regarding Dembski's paper, referenced by 2ndclass above, in his false positive pulsar example, I am trying to see how you reach your conclusion that the pulsar would register as "designed".
quote:
This particular signal repeats every 1.337 seconds.
At 25 bits/s, that's 33.4 bits of information for one period. Since additional periods captured at that rate yield no additional information, there shouldn't be any need to sample for an hour. Now apply Dembski's formula:
SC = -log2( 10^120 * SpecRes(T) * P(T|H) )
I concur with the chance hypothesis being random noise, but that gives us only P(T|H) = 2^34; that might be even less if one compresses out the silence occupying the other 1.297s not occupied by the pulse. Plugging this much in gives us
SC = -log2( 10^120 * SpecRes(T) * 10^10 )
Now, no matter whether SpecRes(T) is 1.7 * 10^10 or smaller, the resulting SC is negative, so the pulsar should not have been considered an indication of design, as I read it. But it all hinges on how one treats the repeating signal initially. Correct me if there is a reason why I shouldn't compress out the clear redundancy in the signal up front.
IP: Logged
|
|
2ndclass
Member
Member # 1979
|
posted 30. July 2006 10:08
GShultz: quote:
At 25 bits/s, that's 33.4 bits of information for one period. Since additional periods captured at that rate yield no additional information, there shouldn't be any need to sample for an hour.
GShultz, under a random noise hypothesis, every sample value is independent and therefore contains additional information.
IP: Logged
|
|
GSchultz
Member
Member # 1953
|
posted 20. August 2006 00:28
Now that you mention it, the circa-1967 pulsar example suggests a trivial, deterministic procedure for converting both random bit strings and ones resulting from physical laws into ones that would register as "design": simply duplicate each sampled bit some fixed number of times. The resulting string would have a much shorter description, but is unlikely to have been generated by a completely random process, so it would register as "design", right? But you can see the flaw: we know that the bits of the latter string are not independent. The appropriate null hypothesis is that each _block_ of bits has a 0.5 probability of being all ones or all zeros.
Would this logic carry over to the sampling of a continuous signal from space? I think it would.
Since the pulses in the example are a real-world signal, we should expect that some sort if analysis will have to be performed on it, over and above simply digitizing it. This may be well outside the scope of the design filter, but so might have been digitizing it in the first place--which has pitfalls beyond what I am discussing here even. Let us assume then that an autocorrelation analysis or a translation into the frequency domain reveals nothing but uncorrelated noise and a prominent, regular "pulse".
Now, if the pulse is coming from a physical object merely obeying the laws of physics, should the null hypothesis be that each state of the object is independent of the previous state? (Dembski is clear that an "appropriate" null hypothesis is needed, not always the random one.) From what we know about the size of celestial objects capable of emitting the amount of energy that we are seeing, nothing would lead us to say with finality that the states of the object are independent. Therefore, it would be inappropriate to employ a null hypothesis that the bits from a sampling are independent. In other words, sampling the signal at a rate far above its obvious period is giving us more information in the raw sense, but not necessarily more independent bits of information. Thus the signal would not trigger a conclusion of "design".
(If Dembski has said that there has to be a _known_ physical process capable of generating a signal before we can write it off as a mere physical process, then that would need some rethinking. The circa-1967 pulsar example does make that clear.)
IP: Logged
|
|
2ndclass
Member
Member # 1979
|
posted 21. August 2006 13:16
GShultz:
quote:
But you can see the flaw: we know that the bits of the latter string are not independent.
The same is true of a Fibonacci sequence, or a long list of primes (which Dembski often uses as an example of manifest design).
The interdependence of the bits is, in fact, essential to the design inference. In Dembski's approach, regularity indicates design, as long as its mechanism is unknown.
quote:
If Dembski has said that there has to be a _known_ physical process capable of generating a signal before we can write it off as a mere physical process, then that would need some rethinking.
Indeed, Dembski insists:
quote:
Indeed, the mere possibility that we might have missed some chance hypothesis is hardly reason to think that such a hypothesis was operating. Nor is it reason to be skeptical of a design inference based on specified complexity. Appealing to the unknown to undercut what we do know is never sound epistemological practice. Sure, we may be wrong. But unknown chance hypotheses (and the unknown material mechanisms that supposedly induce them) have no epistemic force in showing that we are wrong.
IP: Logged
|
|
rolandt
Member
Member # 646
|
posted 01. September 2006 19:36
This is in response to 2nclass's original reply where he feel's he has basically disproved SC in relation to design just because he can find an example of it that does not occur within the context of designer.
This is a problem of most scientific thinkers in that they feel they must map every definition of a contemplation into a mathematical construct. It is as if without mathematics, there is no science. Mathematics is only a formal language which lends itself to descriptive proofs. Which lends itself to science in making conclusions.
But mathematics can be applied to the imaginary or the non-physical just as easily as to the physical, and science deals with only that which is physically observable. Therefore attempting to define many scientific (that which is observable) concepts within a mathematical construct is a mistake because the math doesn't do the whole job.
It's like trying to describe one of Mozart's symphonies in differential equations. It might be able to be done, but the formulas do nothing in portraying the beauty of the music. Just as the music written on paper does nothing for the ears, which is what the music is intended for.
Describing SC with a mathematical construct as Dembski has done is a mistake, because, as 2ndclass has pointed out, he can easily apply the logrithmic equation to a form of SC that has no originating intelligent designer. 2nclass's example is legitimate, by Dembski's definition, and disproves the concept that SC is evidence of an originating intelligent designer.
But the problem is that SC is much more then the mathematical definition provided by Dembski. Take Dembski's definition and try to apply to a Picasso, a Monet, a DaVinnci, Beethoven's complete works, the architectural specifications to Cray's latest supercomputer. It doesn't work. Yet the SC of such things is undeniable.
Specified Complexity is a fact, it is undeniable, and it can be defined, but defining it mathematically is a mistake. To deny specified complexity is not being scientific because it can be observed and therefore identified by almost every human.
To disassociate SC with an intelligent designer is also a denial of the facts. Just as in a mathematical or philosophical proof, it must be considered a given. Because it is difficult or impossible to create mathematical formulas to sufficiently describe SC does not make it non-existent or non-scientific, and therefore cannot be scientifically ignored.
IP: Logged
|
|
|
Printer-friendly view of this topic