|
Author
|
Topic: Exhaustive Sweep of Chance Hypotheses
|
William A. Dembski
Member
Member # 7
|
posted 13. June 2002 17:57
One of the main crticisms of my formulation of specified complexity is that it can't deliver what it promises, namely, an exhaustive sweep of chance hypotheses. Proponents of a purely naturalistic evolution insist that attributing specified complexity to biological organisms is merely an argument from ignorance, closing the door to some chance hypotheses but not the actual one(s) responsible for the biological structure in question.
Against this I urge that there can be independent reasons for thinking that we have a good handle on all the chance hypotheses that might be operating in a given instance. Crucial here are those instances where the natural processes known to operate do not privilege one outcome over another, thus entailing a contingency that unknown laws are not able overturn. I elaborate such an argument in my two responses to Richard Wein on the ISCID archive. In my second response I describe the case of J. Hendrik Schön, which I'll review here briefly.
On May 23rd of this year the New York Times reported on the work of "J. Hendrik Schön, 31, a Bell Labs physicist in Murray Hill, N.J., who has produced an extraordinary body of work in the last two and a half years, including seven articles each in Science and Nature, two of the most prestigious journals."
Schön's career is on the line. Why? According to the New York Times, Schön published "graphs that were nearly identical even though they appeared in different scientific papers and represented data from different devices. In some graphs, even the tiny squiggles that should arise from purely random fluctuations matched exactly." As a consequence, Bell Labs appointed an independent panel to determine whether Schön "improperly manipulat[ed] data in research papers published in prestigious scientific journals."
The theoretical issues raised in this case of putative data falsification are precisely those that my own work on design detection seeks to address. The match between the two graphs in Schön's articles constitutes an independently given pattern or specification. Moreover, the random fluctuations in the graphs are highly improbable. What’s more, the randomness here is well-understood. As a consequence, no unknown mechanism is being sought for how the graphs from independent experiments on independent devices could have exhibited the same pattern of random fluctuations. At issue is the question of data manipulation and design, and we resolve it by identifying specified complexity.
The Schön case seems especially to confirm the ability of specified complexity to provide an exhaustive sweep of chance hypotheses because of the probabilistic independence between random devices (whose randomness is itself well-understood).
It would be good to have a catalogue of examples where the elimination of chance hypotheses by means of specified complexity convincingly constitutes an exhaustive sweep of chance hypotheses as in the Schön case. I'm hoping this thread will generate additional examples of particularly clear cases of such exhaustive sweeps.
IP: Logged
|
|
Danpech
Member
Member # 163
|
posted 13. June 2002 23:32
Is the randomness of the devices the pivotal issue here, or is it the randomness of the objects on which these two devices were presumably used? I was just thinking that perhaps there exist various objects that share a randomness and that, under certain conditions, they exhibit virtually the same patterns.
(I like to leave the benefit of the doubt, just so I can know that I have attended to all logical possibilities---since I'm ignorant of the details of so many things.)
IP: Logged
|
|
fish
Member
Member # 213
|
posted 14. June 2002 11:09
The situation here is quite peculiar. It is the repetition of a particular pattern that allows one to make the inference (non-natural causes), rather the pattern itself. It is far from clear that this case can be thought of as analogous to cases where no such repetition occurs.
This is my interpretation of inference in this case:
The initial hypothesis about the squiggles on the second graph was that they were random.
This situation changed dramatically when the same pattern of squiggles appeared on a previous graph from the same author.
We then have two active hypotheses:
(1) The second graph is random.
(2) The second graph is a copy of the first.
Possibility (2) implys that the author copied the first one - plausible, since there has certainly been opportunity and motivation, although given a low prior probability, since most scientists do not cheat. This hypothesis makes a very specific prediction (that there are elements of the two graphs which are identical), which have to be realised in order for the hypothesis to survive.
Possibility (1) Implies that by chance, the two graphs have the same pattern of squiggles. This is given a high prior probability, since we dont expect scientists to cheat.
If we assume that squiggles take some specified random behavior, then we can calculate the likelihood of the observed pattern under possibility (1). If this likelihood is sufficiently small then possibility (2) finishes up having a higher posterior probability than possibility (1).
There is also a third possibility: that we dont understand the underlying distribution behind the squiggles and that there similarity reflects a different combination of law and chance. However this possibility is not well defined or specific enough to be tested. It is therefore not allowed on to the table.
Nevertheless, the scientist could subsequently clear his name by making some specific and ultimately provable hypothesis of this sort (equivalent to proving that they coin was biased).
I dont see that we have eliminated all "chance hypotheses". We have just eliminated the specific one chance-and law hypothesis that we allow onto the table: that the squiggles are random, with a particular distribution. It is only because this distribution is believed well understood (or in other words that we have no other hypotheses) that we can conclude (2).
Wrt Life, we have no simple, well defined explanation that is comparable to (2), nor is there a single, quantifiable chance-and law hypothesis on the table.
It is all a question of what hypotheses are on the table at a given time.
[ 14 June 2002, 11:47: Message edited by: fish ]
IP: Logged
|
|
warren_bergerson
Member
Member # 262
|
posted 14. June 2002 16:16
On the surface, the concept of ‘a measure of complexity’ which produces an ‘exhaustive sweep of chance hypothesis’, seems both impossible and unnecessary.
It seems to me that the ‘complexity too great to be explained by chance’ is like trying to pick a number bigger than any other number. It is logically impossible. For the sake of discussion, let’s assume that in our universe chance can never produce an phenomena with a probability of less than 1 in 4. We now discover a phenomena with a complexity of 32 which means the probability it arose by chance is 1 in 32.
Have we discovered a phenomena which can not be explained by chance. The obvious answer is ‘not necessarily’. The observed phenomena may be explainable by a series of six deterministic steps each involving chance or stochastic probabilities of 1 in 2. The event arose by chance, in a sense, but none of the probabilities involved was greater than the limit of complexity due to chance.
But if ‘an exhaustive sweep of chance hypothesis’ is logically impossible, it is also unnecessary. The complexity/chance issue being discussed involves, IMO, scientific knowledge, not some abstract issue of absolute knowledge. In terms of scientific knowledge, complexity relative to a specified scientific model or theory. This means that the deterministic processes or mechanisms must be specified before we begin to evaluate the chance or stochastic fluctuations. Furthermore, stochastic fluctuations in scientific analysis are limited to reproducible and testable stochastic fluctuations.
Based on the rules of scientific analysis, and using neo-Darwinian genetics(NDG) to define known deterministic mechanisms and reproducible stochastic it not at all difficult to demonstrate complexity which ‘provides a sweep of chance hypothesis relative to NDG.
In my opinion, the issue here is not whether specified complexity works, but what it is attempting to accomplish. If specified complexity is attempting to establish some type of absolute truth, then the philosophers tell us the effort is doomed to failure. If specified complexity is attempting to establish some type of scientific truth relative to NDG, then it succeeds with huge margins for error.
IP: Logged
|
|
charlie d.
Member
Member # 159
|
posted 14. June 2002 19:10
Let's assume that it will eventually turn out that the 2 graphs are identical because they are indeed duplicates of the the same, but appeared in 2 different contexts because of an accident, e.g. mislabeling of a computer file by a postdoc.
This would mean that the two graphs are not the same by chance as originally intended by Dembski (i.e. 2 machines giving the same noise pattern in 2 independent experiments), but also equally certainly that they are not the same by design, i.e. intentionally.
Now, would this indicate that, alas, Dembski's filter failed to consider one (of the many) chance hypotheses? Or did his filter consider this hypothesis as well, and if so, how?
IP: Logged
|
|
Paul A. Nelson
Member
Member # 26
|
posted 15. June 2002 10:07
Over the next few months, we'll have an opportunity to watch an application of the Explanatory Filter in real time. Objects resembling geometric monoliths have been discovered on the ocean floor (~2,000 feet deep) off the southwest coast of Cuba. Scientists working with the site speculate that the monoliths were not naturally caused, but may be the work of an ancient and previously unknown civilization.
Here's the 5/31/02 National Public Radio story on the discovery:
http://search1.npr.org/opt/collections/torched/atc/data_atc/seg_144240.htm
It's clear that the researchers have already set up "naturally caused" and "intelligently caused" as distinct categories, and the next step in their work is to try to find the diagnostic markers of the second category.
IP: Logged
|
|
Jay
Member
Member # 268
|
posted 15. June 2002 15:18
Great link, Paul! I especially liked the fact that we don't know who did this or how (we don't know the designer yet) and yet we can still make the inference that this may have been designed - all this without even rigorously testing it and ruling out all possible chance hypotheses, and without any indepedent knowledge of the designer.
Now imagine the reaction if I told these guys that they're just trying to play ancient-construction-agents-of-the-gaps in calling this possible design, and then told them that they need to get to work looking for natural methods to produce these geometric shapes! Imagine if I told them that they were not being scientific since they were not looking for repeatable natural laws, and did not rule all possible natural scenarios out! That would never fly. They'd go right on looking for design, and do so in a scientific sense.
When it comes to neutral objects like these shapes, there seems to be no problem in looking for design (and getting funding for it, no less). And no one whines when we do suspect it. But all of that changes when it comes to biology. Here, we hear the cry that although nature *appears* designed, we need to *remind* ourselves that it is not, and that all attempts to infer it are pseudoscience. I think that when we look at arguments criticizing the design inference, we need to compare it to the level of criticism that people like these geometric shape hunters get and ask why there is such a large, glaring disparity.
Anyway, it will be neat to see how this turns out.
Thanks,
jay
IP: Logged
|
|
Evan
Member
Member # 164
|
posted 15. June 2002 15:54
I must admit, both to myself and ID supporters on this Board, that I don't understand how the examples in this thread relate to biology.
In both cases being considered here, the objects under discussion (the graphs and the monoliths) are like things that we know are designed by known intelligent agents - human beings - whose general powers and motivations are well known. So even though the specifics of who did what, when, and why, are subject to investigation, the beginning inference that design is a possibility is predicated on the fact that we know that designers of that kind of thing exist.
So even though some of Dembski's ideas in this regard do seem to apply to such things as codes, forgeries, etc, I don't see the immediate connection between those topics and biology.
I know that such cases are meant to be analogies to the detection of design, but analogies by themselves do not prove anything.
So I have two questions:
a) Are there independent reasons to believe that the analogy between copied graphs or unknown monolithic structures and biology is more than just an analogy, but is rather a legitimate extension of the same idea? and
b) Does the known existence of the types of intelligent agents who could have copied a graph or built a monolith make those situations fundamentally different from biology?
IP: Logged
|
|
Jay
Member
Member # 268
|
posted 15. June 2002 19:24
Evan,
"So even though the specifics of who did what, when, and why, are subject to investigation, the beginning inference that design is a possibility is predicated on the fact that we know that designers of that kind of thing exist."
Well, I don't have a fast enough connection to actually watch this program, but it sounds to me like we have something very unusual - these are really weird geometric shapes, found way, way down under water.
To infer that these underwater things are even human invention is to argue by analogy. We're simply infering that these striking geometric shapes don't fit what we know of natural forces, but instead are better explained as being from design since they also look analogous to other designs that we are aware of. But remember, we're only arguing from analogy here. We don't know who or what the designers are, and have no assurance that they are even human. We must extrapolate from what we do know as designers.
Now let's do this same thing again with the flagellum. We know how to make rotary motors, and are aware of many of the design inputs that are needed to get a good one made and running. We have our analogy. And just as these underwater shapes are analogous to, say, pyramids, so the flagellum is analogous to a rotary motor.
Quite frankly, I think that the flagellum/rotary motor is a much stronger analogy for design than the geometric shapes are to our other geometric shapes that we make. However, as I pointed out earlier, when it comes to biology, people tend to get really squeamish about using analogy for design, even when the analogy is much better than for other things, like geometric shapes under water.
Once again, the scientific community has no problem funding research for chasing after design, even when we don't know who or what the designer is. Heck, look at the Mars stuff. They've been getting false positives for design over and over yet they are still funded and still accepted by the scientific community and still continune to look for traces of nonhuman design - even after their many mistakes. Imagine how someone like Demski would have been treated if he got that many false positives for biology!
For some reason, We are allowed to run full force in our pursuit of traces of design except when it comes to biology. In biology, we are suddenly supposed to take our scientific design detectors off, and remind ourselves that it's all chance+law, no matter how much it looks otherwise. And anyone who who dares pursue design detection here is labeled a charlatan and harshly rebuked (and usually thought to be a closet religious maniac). There is, once again, a clear disparity in criticism.
Thanks,
jay
IP: Logged
|
|
charlie d.
Member
Member # 159
|
posted 15. June 2002 20:28
That is indeed an interesting story. To push the example a little farther, let's assume some signs of manufacturing are found, like chisel marks, maybe some accompanying artifact in their proximity. Clearly, we will then safely infer design. Yet, not knowing what civilization may have created this things, now even whether such a civilization existed in the region, what do people here think the conclusion will be?
1. Despite the evidence for such a civilization, we know humans inhabited this region at the time, therefore a naturalistic explanation should be chosen: some people, unknown to us so far, indeed created these monoliths.
2. In the absence of indipendent evidence for the civilization, a supernatural explanation is preferable: aliens, or divine beings, made the monoliths.
Really, I am curious.
IP: Logged
|
|
Micah Sparacio
Member
Member # 6
|
posted 15. June 2002 20:59
Charlie,
Aliens making the monoliths would still be natural (not supernatural). The dichotomy that the explanatory filter makes is between intelligent and non-intelligent causes (not natural and supernatural): to say otherwise is to not take Dembski's work on its own terms.
I think it is safe to say that THE single most important feature of this find will lie in whether or not it had an intelligent source. After that has been determined, the nuances and details of its source are open for consideration.
IP: Logged
|
|
Evan
Member
Member # 164
|
posted 15. June 2002 23:09
Hello Jay.
I would like to to respond to the substance of your post, but first I want to clarify a few tangential things.
You write,
quote:
For some reason, We are allowed to run full force in our pursuit of traces of design except when it comes to biology. In biology, we are suddenly supposed to take our scientific design detectors off, and remind ourselves that it's all chance+law, no matter how much it looks otherwise. And anyone who who dares pursue design detection here is labeled a charlatan and harshly rebuked (and usually thought to be a closet religious maniac). There is, once again, a clear disparity in criticism.
In my defense, I don’t believe I said anything to imply that anyone who looks for design is a “closet religious maniac”, or a “charlatan,” and I don’t believe I rebuked anyone. I have asked some questions about whether design detection in such fields as archaeology or forensics science can be extrapolated to biology, and I think that is a reasonable topic to explore.
Secondly, and even more parenthetically, you write
quote:
Heck, look at the Mars stuff. They've been getting false positives for design over and over yet they are still funded and still accepted by the scientific community and still continue to look for traces of nonhuman design - even after their many mistakes.
I wonder what you are referring to here? Of course, at one time they thought there might be canals on Mars, but that was a long time ago. Currently, there is interest in looking for life on Mars, but that is different than looking for designed objects. So I really have no idea what “many mistakes” you could be talking about, and I am curious if what they might be.
So now to the main points.
I asked two questions previously, which I will restate here more directly:
1) Analogies by themselves prove nothing. They may help us pursue a line of reasoning, and therefore motivate a line of research. But given that analogies are by definition imperfect (for if they were not they would be equivalencies), the question of whether any part of an analogy holds must be investigated empirically. So the fact that the flagellum appears to be analogous to a rotary motor might cause one to suspect design, but that it is all it can do. The analogy itself is not in any way evidence that the flagellum is in fact designed - that has to be investigated directly.
To give a few superficial examples, river systems are like city sewer systems, “buckyball” molecules are like soccer balls, snowflakes are like geometric patterns we draw on paper, and photons shot from a light source are like bullets (under some conditions.) And yet in all these cases, design is not inferred; and in all these cases the ways that the analogy does not hold is one of the considerations.
So the fact that something is analogous to something we make is not, in and of itself, evidence for claiming design.
Let me make it clear that I know that other important criteria for inferring design are explained by Dembski, such as whether it is probable that natural forces were the cause of an object’s existence - these are what need to be investigated.
2) We know that human beings exist, and we ascribe intelligent agency to them. We know roughly what our abilities and motivations are. Therefore, when we apply design inference theory to forensic or archaeological situations, we have a two-fold line of reasoning: both the positive evidence that this is like something that we know human-being-type creatures can design and manufacture, and the negative evidence that we believe that natural forces are extremely unlikely to have have caused the situation in question.
However, in biology, we only have the negative evidence, because we know of no intelligent agent analogous to human beings with the ability to have been busy creating certain aspects of life for billions of years. We might think we can infer the existence of such an entity based on analogy with human beings, but that inference is subject to the flaws I pointed out in part 1: analogies don’t prove anything by themselves.
================================================
I want to again make it clear that my point here is not to dismiss ID, but rather to state that I don’t think that analogies from engineering, archaeology, or forensics science actually amount to any evidence whatsoever for design in biology.
Design in biology must be approached on its own merits. The ways that biology is different from those analogous things is much more significant than the ways it is the same. If ID is going to be established in biology, it needs to be done by direct reference to biology itself, not by analogy with machines, forgeries, or archaeological artifacts.
IP: Logged
|
|
Paul A. Nelson
Member
Member # 26
|
posted 16. June 2002 10:12
Evan,
Figuring out that a monolith (or assemblage of monoliths, such as Stonehenge) was intelligently caused is not analogous to a design inference. It is a design inference. If the Bell Labs researcher, J. Hendrik Schon, currently under suspicion for data falsification, is found guilty, it will be on the grounds of a design inference. Not an analogy to a design inference.
The point of analyzing design inferences in general, or as a type of inference, is to see what we can learn about the logical and evidential structure of such arguments. Of course biology raises deeper challenges than Stonehenge. But I disagree with you that there is something different in kind about biological objects. And engineering metaphors can be profoundly illuminating in biology: Consider "transcription," "translation," "reading frame," and so on.
[ 16 June 2002, 10:26: Message edited by: Paul A. Nelson ]
IP: Logged
|
|
charlie d.
Member
Member # 159
|
posted 16. June 2002 11:03
Micah:
let's say "preter-natural" then, since the existence of aliens and their ability to reach Earth, you will agree, have not been demonstrated yet.
My question stands: would you (and the others here) opt for a seemingly unlikely, but still naturalistic explanation, or for an at least equally unlikely, preter-natural explanation, which appeals to agents for which we have no evidence for?
There is really no wiggle room here, and of course the answer that is given (not by us, of course, but by archeologists/underwater explorers and funding agencies) would have a significant impact on the next step of the research on this matter.
[PS: I actually still think my first question was more relevant to the initial point, and I'd like to reiterate it: if it is shown that Schon's duplicated plots are due to human mistake, rather than willful deceit, would that count as a "chance" event (I think it should), and if so, how would that be included in the calculation of chance hypotheses? Of course, if the calculation is as "exhaustive" as claimed, it must include a specific step which would account for this kind of occurrance. I'd love to hear any comments on this.]
[ 16 June 2002, 13:35: Message edited by: charlie d. ]
IP: Logged
|
|
charlie d.
Member
Member # 159
|
posted 16. June 2002 11:08
Paul:
if Schon is found guilty, I am quite sure it is going to be based on some sort of additional empirical evidence, not only on a statistical inference. (Incidentally, I don't even know how one would actually calculate the chances of that background noise being the same by chance - do you?)
IP: Logged
|
|
|
Printer-friendly view of this topic