|
Author
|
Topic: Murder by Design?
|
Frances
Member
Member # 169
|
posted 26. September 2002 12:26
As far as Dembski's claim of a design inference related to the Bell Labs investigator I am somewhat confused what relevance this has to the design inference ala Dembski?
In many instances there could be instances of chance (sloppiness in using graphs) explaining the coincidences found in these graphs. So was the design inference intelligent design or was it a chance hypothesis? In fact the example has even less applicability to show that the DI is useful for biology where in addition to chance hypotheses also regularity hypotheses exist.
Since the original data have not be retained, the committee reaches a conclusion that
quote:
At a minimum, Hendrik Schön showed reckless disregard for the sanctity of data in the value system of science. His failure to retain primary data files compounds the problem.
Surely not a design inference since Schon argues that the similarities are due to honest mistakes.
Without the admissions of Schon the committee would have no real data to conclude a design inference in most of the cases.
In conclusion: While this example is surely not very convincing it is even less convincing that the DI has any relevance when in addition to chance hypotheses there are regularity hypotheses to be considered.
IP: Logged
|
|
Mike Baughman
Member
Member # 270
|
posted 26. September 2002 13:55
Earlier I said the name is irrelevant to the probabilities and only supplies the specification. On further thought I think this is incorrect. The name would be relevant to the numerator, i.e. we want to find P(E) where E = "a Harris County resident named Mary Morris is murdered in a 3 day period".
I made a very rough but conservative estimate (using the Chicago phone book) and got a factor of 1.9e-5 for the name "M Morris" or "Mary Morris". Using the over-18 population of Harris Co. of approx. 2.416 million yields 45 Mary Morrises. So P(E) is 6.33e-7 * 45 = 2.85e-5 and the probability of two random Mary Morris murders in 3 days is 8.11e-10. Factoring in time of day (after dark) and MO (left in car, no robbery) would further reduce this number.
I used a Probability Bound of 7.9e-10 (based on the population of Harris Co. times days in a year) and concluded that the real Mary Morris murders are right on the borderline (coincidence is just barely possible) while the hypothetical Mary Smith or Wanda Z. Kowzlowski murders are definitely not coincidence. Of course this depends on my PB; what do people think is a reasonable PB?
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 26. September 2002 14:07
O.K., how do I straighten this out gracefully? I got myself cross-threaded, which I thought only happened to nuts. ![[Smile]](smile.gif) My remarks addressed to Mike Gene in the post above belong in the "Does Darwin Predict ..." thread, and I've moved them over there, editing them out of the post here. The MDT remarks and the comment on the Schon posting belong here, so I'll leave them. Now it's time to walk the dogs!
Sorry for the confusion.
RBH
IP: Logged
|
|
Mike Baughman
Member
Member # 270
|
posted 26. September 2002 14:19
About the Schoen case, Dr. Dembski asks, "Is there 'actual evidence' of data falsification?"
The Committee's answer for 16 of 24 allegations is "yes." quote:
In the end, the Committee concluded that, of the 24 Final Allegations examined, Hendrik Schön committed scientific misconduct in 16, some of which were interrelated. Of the remaining 8, 2 were judged to have no clear relationship to publications, while 6 were troubling but did not provide compelling evidence of scientific misconduct.
IP: Logged
|
|
charlie d.
Member
Member # 159
|
posted 27. September 2002 09:57
From quickly reading the executive summary , I would say the committee concludes misconduct on these bases:
- Schon showed appalling disregard for common scientific practices (good recordkeeping, preservation of original data)
- Schon admitted misrepresenting data in several occasions (e.g., using functions in plots instead of real data points without clearly stating it)
- Schon admitted reckless error in several occasions ("mistakenly" showing the same plot in in different contexts).
This already fully fits the definition of scientific misconduct and would be more than enough to "bury" any scientist.
As for inferring falsification in the cases not admitted by Schon, as far as I can tell the committee basically says they are as sure as one can humanly be without being able to see the actual data (because Schon deleted it from his computer, or discarded it), and actually proving it.
Either way, I think this is very different from a purely eliminative, ID-like design inference in a biological context: the committee has very clear "models" of who could have falsified data, and how (indeed, they worked hard to identify, and describe in detail, several of the likely "mechanisms": pasting curves from different experiments together, applying simple functions to shift curves up and down, etc, and they stress repeatedly that Schon alone witnessed the experiments, and collected and organized the data for the papers).
Can anyone even imagine the committee report saying, NFL-style:
"We have no clue about how any known material agent (Schon or someone else) could have falsified the data through identifiable material mechanisms, nor we feel this is necessary. Because we can't explain how the data arose, we infer falsification"?
[ 27. September 2002, 10:00: Message edited by: charlie d. ]
IP: Logged
|
|
Mike Baughman
Member
Member # 270
|
posted 27. September 2002 11:12
Charlie,
You have mis-characterized the inference. The operative principle in the Schoen case and the murder cases is eliminating chance. The Schoen data was initially distrusted on the basis of CSI: chance is not a viable explanation for a highly improbable, specified event. The specific mechanism isn't necessary to the inference of "not a coincidence".
IP: Logged
|
|
Frances
Member
Member # 169
|
posted 27. September 2002 12:45
Mike
Seems that CSI may have some usefulness if regularity can be eliminated totally before applying the filter. Of course with evolution it is in fact the combination of chance and regularity which is argued to be able to generate CSI. As for instance Wesley and others have argued, the design inference may have some usefulness as it has been already applied for ages to archaeology, criminology for instance but its usefulness in other sciences has yet to be shown applicable.
IP: Logged
|
|
charlie d.
Member
Member # 159
|
posted 27. September 2002 18:29
Mike:
sure the committee eliminated chance, but it did so because it had the pretty obvious, much more likely alternative of a known material agent acting through known material mechanisms.
However, imagine now that two identical sets of graphs, equally improbable by chance as Schon's, had appeared in separate papers, from independent laboratories, working contemporaneously in different parts of the world and demonstrably unknowingly of each other. The probability of the two sets of graphs appearing by chance would be exactly the same as Schon's, but would you infer design in this case, in the absence of an obvious known material mechanism to generate the replicate graphs, or would you claim ESP or something? Do you think the committee would have?
Inferring design is possible only when intelligent agents are known to operate. Otherwise, coincidences are just coincidences (at least as a default explanation, until a likely causal mechanism is identified).
[ 27. September 2002, 18:30: Message edited by: charlie d. ]
IP: Logged
|
|
Mike Baughman
Member
Member # 270
|
posted 28. September 2002 14:00
quote:
Posted by Charlie D.:
However, imagine now that two identical sets of graphs, equally improbable by chance as Schon's, had appeared in separate papers, from independent laboratories, working contemporaneously in different parts of the world and demonstrably unknowingly of each other. The probability of the two sets of graphs appearing by chance would be exactly the same as Schon's, but would you infer design in this case, in the absence of an obvious known material mechanism to generate the replicate graphs, or would you claim ESP or something? Do you think the committee would have?
Hi Charlie,
If the situation imagined were to actually occur, I think the proper conclusion would be that the results are not coincidentally identical. I would not attribute it to ESP but would say that "some mechanism as yet unknown" had operated. I don't believe that this unknown mechanism should be ruled out a priori. I think that if a committee wrote off their results as coincidence that the ruling would be challenged on the basis of TDI.
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 28. September 2002 15:30
Mike Baughman wrote: quote:
quote:
quote:Posted by Charlie D.:
However, imagine now that two identical sets of graphs, equally improbable by chance as Schon's, had appeared in separate papers, from independent laboratories, working contemporaneously in different parts of the world and demonstrably unknowingly of each other. The probability of the two sets of graphs appearing by chance would be exactly the same as Schon's, but would you infer design in this case, in the absence of an obvious known material mechanism to generate the replicate graphs, or would you claim ESP or something? Do you think the committee would have?
Hi Charlie,
If the situation imagined were to actually occur, I think the proper conclusion would be that the results are not coincidentally identical. I would not attribute it to ESP but would say that "some mechanism as yet unknown" had operated. I don't believe that this unknown mechanism should be ruled out a priori. I think that if a committee wrote off their results as coincidence that the ruling would be challenged on the basis of TDI.
There's one real straightforward explanation for the case charlie proposes: The two laboratories were both investigating the same phenomenon in the same way, using a research paradigm in which an experimental manipulation generates 'canalized' effects, possibly via some dynamical pathway tightly constrained by purely natural means (say, constraints imposed by a tightly controlled physical environment) so that the time course of the system is highly replicable. In that instance, nearly identical graphs of the temporal course of the phenomenon under study would be produced by purely natural means without invoking ESP, CSI, or probabilistic handwaving. Replication, after all, is an important component of scientific research. Since labs are regularly replicating each other's work, the set of potential matches is relatively tightly constrained and probability estimates over some hypothetical reference universe of 'all possible curves' are simply invalid and irrelevant. Elimination of the 'regularity' alternative is not as easy or straightforward as IDists seem to think, and require more than casual knowledge of the phenomenon under study. That doesn't come from reading secondary sources intended for a lay audience.
RBH
IP: Logged
|
|
charlie d.
Member
Member # 159
|
posted 28. September 2002 16:23
RBH:
Of course reproducibility, to a point, is expected, but I was actually referring to the graphs that showed identical (expectedly random) "noise" in different experiments, and to different experiments reading different variables with different equipment giving the exact same curves. I can't think of any natural law that would replicate that, especially if it can't be replicated elsewhere.
Mike:
What kind of a "yet unknown" mechanism do you have in mind? Certainly, you would not claim intelligent design, would you? Maybe, like RBH, you could claim some yet unknown natural mechanism, regularity or law, but not design, I am sure. Indeed, in this context, design would be by definition a supernatural explanation, like ESP (because we know of no natural designing intelligence that could accomplish that kind of simultaneous graph "trick"). So, your design inference here would not be a design inference at all, just a chance elimination process (which is what Dembski's method really is, btw). In the absence of a likely material intelligence, one would claim coincidence or regularity, even of an unknown sort.
Here's another good one for TDI:
quote:
Reuters
March 6, 2002
Twins Killed in Identical Road Accidents
HELSINKI (Reuters) - Finnish twin brothers, aged 71, were killed in
identical bicycle accidents along the same road two hours apart, police
said Wednesday.
"This is simply a historic coincidence. Although the road is a busy one,
accidents don't occur every day," police officer Marja-Leena Huhtala told
Reuters.
"It made my hair stand on end when I heard the two were brothers, and
identical twins at that. It came to mind that perhaps someone from
upstairs had a say in this," she said.
One twin was hit by a truck and killed while out cycling early Tuesday on
the west coast of Finland.
Before police had identified the body and informed family members, his
brother was killed on his bicycle by a second truck a half-mile down the
road.
Note that it turns out that the twins had no reason to commit suicide, and they were both reportedly in excellent moods before setting off for their fatal bike rides.
The police ruled coincidence, despite the "DI-friendly" quote from the policewoman in the text. Would you?
[ 28. September 2002, 16:24: Message edited by: charlie d. ]
IP: Logged
|
|
Mike Baughman
Member
Member # 270
|
posted 30. September 2002 09:56
Hello RBH,
What Charlie said. (re identical results - see his reply).
Charlie,
What kind of mechanism would you propose? I'm not trying to be evasive, but your hypothetical is exactly identical results, which as you say cannot be explained by any known ordinary naturalistic process - so by your own definition either chance or some non-naturalistic or unknown natural cause must therefore be at work. I merely assert that in this situation one of the latter two is more likely than chance - given enough and long enough curves, chance could be definitely ruled out.
This is all a fantasy experiment, so why would I not entertain the idea of some sort of intelligence at work? Why must I rule it out except as a matter of personal a priori philosophical inclination?
As for the Finnish twins, it may not be that huge a coincidence. Depends on the amount of traffic, how dangerous that stretch of road is, how often the twins rode their bikes over that stretch, etc. Traffic deaths are more common than murders, and the twins apparently had similar lifestyles. I imagine the calculated probability would be well within any applicable Probability Bound, certainly Dembski's UPB.
BTW I have asked above and will repeat the question, because I really am curious - what do people think is a reasonable PB for the Mary Morrises?
IP: Logged
|
|
charlie d.
Member
Member # 159
|
posted 30. September 2002 13:22
quote:
I'm not trying to be evasive, but your hypothetical is exactly identical results, which as you say cannot be explained by any known ordinary naturalistic process - so by your own definition either chance or some non-naturalistic or unknown natural cause must therefore be at work. I merely assert that in this situation one of the latter two is more likely than chance - given enough and long enough curves, chance could be definitely ruled out.
Oh, I agree you can reasonably rule out chance in the case (in fact, I do not think a case like that will ever happen, improbable as it is). The point is, however, that just because you rule out chance, you can't automatically infer design, like Dembski would want us to believe. The Schon example is simply a case in point: in the real affair, we statistically rule out chance and infer design because we know a designing agent could have acted through material mechanisms. However, take away the knowledge of a likely designer or design mechanism, as in my imaginary Schon-less case, and one is stuck with saying that one simply doesn't know how it happened.
I see you also - correctly, IMO - fall short of inferring design simply based on chance elimination in the imaginary case. In fact, I am sure if you had a plausible material explanation for that improbable result (e.g., the noise readings may have been influenced by cosmic radiation from a solar storm while the samples were being read), you would possibly mention that as the most likely possibility, rather than a supernatural one like ESP, even in the absence of "a detailed, step-by-step explanation" (which is what Demsbki is asking for the flagellum).
So, basically, the claim that the Schon case is a) a design inference entirely based on the elimination of chance possibilities, and b) a valid analogy for the case of the flagellum, does not stand up to scrutiny. People do NOT normally, and in fact SHOULD NOT infer design simply based on statistical improbability considerations, in the absence of a likely designer and/or mechanism of design (and we do not have either for the flagellum). Moreover, nobody claims the flagellum arose by chance alone, so elimination of chance in its origin is really nothing new or noteworthy.
[ 30. September 2002, 16:36: Message edited by: charlie d. ]
IP: Logged
|
|
Mike Baughman
Member
Member # 270
|
posted 01. October 2002 11:35
Hi again Charlie,
As far as I can tell, the Schoen illustration is not being claimed as anything but an instance of "elimination of chance" on the principles of TDI. It is an illustration of one application outside of the biological design inference.
Question for Dr. Dembski or anyone who'd like to take a stab at it: how does one determine the Probability Bound for a specific application? I realize the UPB was chosen so as to effectively eliminate all false positives; but it does so at the cost of many false negatives. Is selecting a PB more or less arbitrary?
IP: Logged
|
|
charlie d.
Member
Member # 159
|
posted 01. October 2002 12:40
quote:
As far as I can tell, the Schoen illustration is not being claimed as anything but an instance of "elimination of chance" on the principles of TDI. It is an illustration of one application outside of the biological design inference.
Maybe I am misunderstanding you. Didn't you say before that the same statistical considerations as in the Schon case, when applied to the imaginary Schonless case, cannot lead automatically to infer intelligent design? That in the Schonless case elimination of chance by statistical inferences (the EF) cannot by itself exclude the possibility of some unknown non-intelligent (naturalistic) mechanism?
What is the difference between the two cases and their inferences, then, but the existence in the Schon case of a known potential designer/mechanism, and absence thereof in the Schonless case? What good is a Design Inference protocol that only allows to infer Design when there is a known potential Designer around?
IP: Logged
|
|
|
Printer-friendly view of this topic