|
Author
|
Topic: Future directions for ID
|
Frances
Member
Member # 169
|
posted 02. November 2002 13:22
I have taken the moderator's suggestions to heart and have pruned my original posting in response to Dembski's paper to maintain focus. While I politely disagree with the moderator's portrayal of my posting(s), I , as a guest to Brainstorms, will abide by his decisions.
I have also rewritten certain parts of what was a response which grew over time to take into account my additional research. I shall attempt to formulate my thesis first and will then proceed to comment on some of Dembski's statements in the hope to reconcile my thesis with the future of ID. Additionally I will attempt to contribute to the scientific future of ID by proposing predictions and explanations of ID. In order to avoid the 'strawman' argument I have provided for links to the source materials. Feel free to add or remove from this list of predictions. This is still in draft format and I am inviting, no encouraging, Brainstorm participants to provide their feedback, positive AND negative.
Thesis:
ID in my opinion can take on two (scientific) forms:
1) a method of providing criticism of present (naturalistic) hypotheses and theories (Darwinism, Neo-Darwinism etc)
2) a method of providing testable and falsifiable alternatives for these hypotheses
The first form is where ID seems to have focused its present attentions. Dembski is wondering how to move ID forward scientifically and I argue that the next step has to be providing alternatives, predictions and explanations. After all the first form is well captured by the present scientific paradigm and in fact Dembski himself shows how present researchers have contributed alternative and controversial hypotheses. Dembski mentions Gould and Margulis but in fact the list of names seems to be much larger than this. Science after all thrives in the presence of disagreements.
Dembski argues that the two principles which drive design are
1. Liberation from ideologies that suffocate the human spirit such as reductionism and materialism.
2. Fresh insights into biology
Without conceding Dembski the first principle, I would like to focus mainly onto the second argument namely that "ID provides us with fresh insighs into biology". Are these fresh insights with respect to the first scientific form or the second scientific form? I would argue that most of the fresh insight so far seems to focus on the first form. For instance, the hypothesis of "Irreducible Complexity I" is focused around showing that "Darwinian theory cannot explain Irreducibly Complex systems", "Irreducible Complexity II", as proposed by Dembski in NFL makes three major changes to IC. The most relevant one for this discussion is to change the function of the IC core which now has to be an original one. Both the original IC (IC-I)as well as IC-II focus around Darwinian mechanisms being unable to generate IC systems. But the argument is a probabilistic one in that indirect pathways are considered possible but unlikely. How are we going to establish if these indirect pathways are indeed more unlikely than its alternatives? One may argue that Behe's arguments have generated renewed interest in systems like the flagellum or ICness and 'Brainstorms" has witnessed a very lively discussion between Dembski and Deanne Taylor on Deanne's hypothesis that for instance gene duplication and variation may be sufficient in generating IC systems. Scale-free or scale-invariance seems to be a property that is found in many biological systems. How are we going to test this hypothesis? So far it seems to generate the same features as found in nature. Is this hypothesis a likely candidate? I would argue yes, so how could ID approach this issue?
I see three options:
1. Eliminate the hypothesis (scale invariance, gene duplication)
2. Propose an alternative naturalistic hypothesis which performs better
3. Propose an alternative ID hypothesis
But 1. and 2. seem to be standard practice in the present scientific paradigm and thus would not provide the 'fresh insight' into biology that is argued to be one of the principles which drive intelligent design. Does this only leave option 3? That depends on the validity of my argument that there are only 3 such options. Perhaps I am missing an option and I am open to arguments which show that I have done so. One may argue that the 4th option would be the design inference filter as proposed by Dembski. Let's explore this option:
Detecting design: The design inference
One of the most interesting approaches of the ID theory is the attempt to propose a way to reliably infer intelligent design (with reliable I mean 'no false positives'). As I have shown elsewhere, the EF may not be the best filter, Elsberry and Wilkins have shown how there is an inherent difference between rarefied design and 'design' which may complicate the design inference. They also noted the role our ignorance plays in the design infererence. But lets for the moment stay with the filter as proposed by Dembski.
Thus we get to the 'reliability' claim. Let's assume that Dembski were correct in his claim that the filter knows no false positives. What would this mean for the design inference? I would argue that it would make the design inference unfalsifiable. There is not tentative conclusion that can be strenghtened or weakened by knew knowledge or insight. It is an 'all or nothing' proposition. But what if it can be shown that there exists a natural path after all? One which was not obvious to us initially. Would this disprove the design inference? I would argue that it would not, it would at most argue against a design intervention but it would not disprove the possibility of 'front loading' or 'deck stacking'. In fact Dembski seems to argue for as much when discussing the Oklo reactor.
What may be the solution to this? We could reject 'deck stacking' as a valid form of intelligent design after all for all practical purposes it can not be distinguished from naturalistic pathways. Does this mean that intervention is the only remaining form of intelligent design which may be falsifiable? Even if we were to argue that this is the case, we have now a new problem to deal with: the separation of 'front loading' or 'deck stacking' from 'intervention'.
Alternatively one may argue against the 'reliability' claim, perhaps one could argue that it is reliable with respect to our present knowledge or that it may be 'unrealiable' under certain circumstances but then we enter the realm of Sober's argument. As long as something is 'reliable' in the sense of no false positives then one could use logical complements to argue for the existence of something by showing the absence of something else. But now this 'something' is not a logical complement as much as a probabilistic complement. There is no certainty that the absence of natural hypotheses identifies intelligent design but rather a probability. And what is this probability? We don't know, we have no reliable way to determine such probability when we argue from elimination. It may be large, it may be small. All we know it exists. The moment we allow a probability to enter the design inference in this manner we have basically allowed it to defeat itself. So this means we have to reject the idea that the filter may be unreliable but that effectively argues in favor of our omniscience or against our ignorance. "No false positives" suggests either omniscience on our part or magic on the part of the mechanism of ID. Omniscience would allow us to determine if we had indeed eliminated all hypotheses, 'magic' would mean that we could eliminate in one big swoop chance/regularity.
Thus my suggestion that: An eliminative inference with incomplete knowledge requires a Bayesian approach.
Biological information and complexity are interesting and hot topics. And despite the claims that intelligence is a requirement for biological information, there seems to be quite a bit of theoretical and experimental data that suggest that information in the genome, in the Shannon sense at least, can increase. The 'Maxwell demon' argument of Adami, Schneider and others shows how the environment can impart information onto the genome.
Dembski continues his argument with the suggestion that ID's legitimacy hinges on two facts. The first 'fact' is that evolutionary biology is claimed to have been hugely unsuccessful as a scientific theory in accounting for the origin of life and the emergence of biological complexity , secondly ID is offered as the 'only alternative to a mechanistic evolutionary biology'.
Lets explore both claims.
Has evolutionary biology been 'hugely unsuccesful' in a) origins of life b) emergence of biological complexity?
Let me first point out that 'origins of life' and 'evolutionary biology' may be two totally different topics. Evolutionary biology deals with what happened to life ONCE it arose. This seems to be a common confusion but its quite important. How do we determine that evolutionary biology has been unsuccesful in the area of origins of life? Certainly the number of hypotheses would indicate that origins of life is a fast 'evolving' research topic, but perhaps hopelessly frustrated by the lack of direct evidence. On the other hand, how succesful has ID been in this area? Origins of life research is still an emerging science but its hypotheses have made quite some inway into our understanding of pathways. After all, one should not jump from the assertion of 'hugely unsuccesful' to the idea that thus ID has done better? Secondly, has evolutionary biology been unsuccesful in explaining the emergence of complexity? I have quoted several very relevant papers in this area which suggest otherwise (Schneider, Adami). More recently Deanne Taylor showed how the concept of gene duplication seemed to explain quite well the observed 'scale free' nature of biology features. Do we know all the answers, will we even know all the answers? I doubt it but how can we determine if ID provides for a better alternative if there are no tools provided to evaluate how well ID performs? For instance does ID explain the scale free nature of so many biological features?
Now we get to the interesting part. In order to further my ideas that ID needs to put itself at risk, that ID needs to propose hypotheses that make predictions and provide for explanations I have been collecting examples of such. Firs of all I would like to identify these predictions and explanations, then I would like to determine if the ID prediction/explanation follows in a logical and defensible manner from the theory of ID and finally I would like to propose to test the ID predictions/explanations against alternative hypotheses.
Predictions of ID
A theory of science needs to place it self at risk. It needs to predict, it needs to explain how it can be falsified, it needs to constrained in its capabilities for it to be an explanation. Dembski argues that ID concedes predictability but I would argue that ID does make predictions. Or at least I argue that predictions based on ID have been proposed.
1. "If these systems are designed, we can expect the information to be densely packed and multi-layered (save where natural forces have attenuated the information). Dense, multi-layered embedding of information is a prediction of ID.
Source
2. "Intelligent design means that life began abruptly", "Fossil forms will appear suddenly and without any precursors"
Source and Source
3. High information content machine-like irreducible complex structures will be found.
Source
4. Genes and functional parts will re re-used in different unrelated organisms
Source
5. TRIZ: While TRIZ does incorporate Darwinian mechanisms it also argues that for true innovation a non-Darwinian approach is required. Source
6. intelligent design would predict the same type of development patterns we see produced by human engineers. These include; 1. A discontinuous or "Quantum" increase in genetic information; 2. A persistence of morphological isolation and stasis and; 3. A hierarchical "top down" pattern of diversification. All of these predictions have analogies in the human design realm. Source
7. This strong proscriptive claim, that natural causes can only transmit CSI but never originate it, I call the Law of Conservation of Information.
Source
Explanations of ID
Some argue that ID makes for better explanations than its alternatives. The following lists a few examples. I intend to extend the list with any new examples I can find. Please feel free to contribute to the list.
1. ID provides a better explanation of the origin of "information", in particular the origin of DNA, than does evolution.
Source
2. ID provides a better explanation of the Cambrian Explosion - the sudden appearance of new phyla in the fossil record 570 million years ago - because new organisms require a new information code.
Source and Source
Critiques of ID
Several authors[1] have written excellent critiques on intelligent design, 'No Free Lunch", the design inference and many of their criticism has remained unaddressed in any satisfying manner. ID needs to engage scientific criticism with scientific response. Issues that remain so far unresolved include:
1. Is the design inference free of 'false positives'?
2. Is CSI calculated with respect to the actual hypothesis being tested or with respect to a uniform chance hypothesis?
3. Can (indirect) pathways as proposed by various authors lead in principle at least to IC systems?
4. If algorithms can generate information why can nature using the same algorithms not generate information?
5. Are there any examples of CSI in nature?
6. How does the design inference deal with incomplete knowledge?
7. Given the complexity of the pathways involved in the flagellum and given the lack of causal history, how does ID propose to formulate probability measures for these real life examples?
The last one is a severe problem since I doubt at this moment that the design inference can be succesfully applied at something more complicated than 'cheating'. After all elimination of pure chance is quite a bit simpler than elimination of regularity/chance combined.
Testing ID inferences
It is not good enough to claim that the design inference is free of false positives, that ID can be reliably infered. We need to put the design inference to actual tests [2]. So far few if any tests seem to have been performed to establish the accuracy of the design inference, the extent of problems to which the design inference can be succesfully applied, comparison of design inference approaches with approaches in criminology (motives, means, opportunity, eye-witnesses, evidence), SETI, archaeology.
Perhaps what we need is a double blind test of the design inference as applied to a variety of problems.
[1] Elsberry, Wilkins, Wein, Sober, Murray, Chiprout and many others.
[2] "Do the calculation. Take the numbers seriously. See if the underlying probabilities really are small enough to yield design." W. A. Dembski
Conclusion
I do not believe that the 'average scientist' would object to ID if it were to formulate some hypotheses which could be evaluated. IF however ID's contribution to science is pointing out weaknesses in present theory then it behaves just as science does. One may be tempted to use the conclusion that criticism of evolutionary pathways is evidence FOR ID but that seems to be begging the question.
So what does ID predict? A curriculum of ID should not merely be pointing out problems in present theory, after all science itself seems to be doing a good job at that (Gould, Margulis per Dembski's own arguments) so do we need an ID curriculum that shows problems or proposes solutions? It's easy to find fault it's much harder to propose improvements. Until ID can rip itself away from anti-Darwinism/anti-methodological naturalism/anti-xyx and become pro-ID, it will keep have to defend itself against the absence of any positive contributions to our knowledge.
Note that I am not arguing against exposing weaknesses in present hypotheses, I believe this to be the foundation of scientific inquiry.
[ 02. November 2002, 13:44: Message edited by: Frances ]
IP: Logged
|
|
Paul A. Nelson
Member
Member # 26
|
posted 02. November 2002 14:34
Great post, Frances. You raise many interesting and important questions.
Let me respond to just one for the moment. You write:
6. How does the design inference deal with incomplete knowledge?
Can you give an example of any science that has "complete" knowledge?
IP: Logged
|
|
Jack
Member
Member # 265
|
posted 02. November 2002 15:28
Frances>>Incomplete knowledge means that we cannot assume that EF has no false positives.<<
So what? Where does evolutionary science take into account the cases of false positives, namely, things thought to have evolved that in fact did not evolve?
False positives in ID research are only a caveat given that ID is a provisional and self-correcting methodology. In fact, if ID is built upon nothing more than false positives, it will
eventually collapse as more information comes in. On the other hand, as more information comes in the ID inference may be strengthened. Yet in the end, something may be a false positive simply because we are dealing with a natural origin for which all the evidence was lost. But this is no different from science scoring a design event as an evolutionary event because we don't have the *type* of evidence science needs.
[ 02. November 2002, 15:58: Message edited by: Jack ]
IP: Logged
|
|
Frances
Member
Member # 169
|
posted 02. November 2002 16:31
I hope that the moderator will allow me to respond to Paul.
quote:
Can you give an example of any science that has "complete" knowledge?
Exactly, I cannot which leads me to the question how ID will attempt to deal with incomplete knowledge. More specifically, how will incomplete knowledge affect the elimination filter proposed by Dembski? I see various ways:
1. Incomplete knowledge means that we cannot assume that EF has no false positives
2. Incomplete knowledge means that we need to test hypotheses side by side to determine which one the hypotheses is the best one, given our present state of knowledge
3. Incomplete knowledge means that we need to provide for a category 'ignorance/we don't know' in the EF not dissimilar from Wilkins/Elsberry's 'improved EF filter'
Incomplete knowledge and eliminative induction almost inevitably seems to require a Bayesian approach. But I am also interested in hearing how Fisherian approaches could be used.
As far as science is concerned, incomplete knowledge is what makes the hypothesis approach so fruitful, it captures 'the best explanation' given our present knowledge without any pretense that this hypothesis will always remain the best. In fact science will attempt to deal with their ignorance to determine if given new knowledge through research, the present hypothesis still holds.
It's a trial and error process in which incomplete knowledge is taken as a given. However if we want to eliminate all hypotheses we can at best claim that we have eliminated known hypotheses. How can we propose a reliable way to infer design?
1. Make design a tentative inference: But why would design be a prefered inference over 'we don't know'? That is not directly obvious
2. Propose a design hypothesis which can be compared to candidate hypotheses.
Perhaps there are additional ways to infer design that deal with incomplete knowledge. Perhaps ID is infered wrt to our present knowledge? But that would make it vulnerable to become a 'God of the gaps' argument, a placeholder until something better comes around. In the end, if ID wants to avoid this, it will need to propose hypotheses that can be compared with competitor hypotheses.
As far as false positives are concerned in science. As I have argued, science embraces the possibility of false positives, namely through the method of falsification. Science it tentative, the best explanation given our present knowledge and data.
Jack:
"False positives in ID research are only a caveat given that ID is a provisional and self-correcting methodology."
Perhaps that is what Jack perceives ID to be but that is not how Dembski has proposed ID toi be infered. ID, as argued by Dembski, is not tentatively infered but is 'reliably' infered where reliable refers to 'no false positives'. Thus it is argued that IF design is infered it is CORRECTLY infered.
THus I would argue that ID is NOT a provisional methodology.
If Jack has a version of ID methodology in mind that does not suffer from this problem then I would encourage him to describe this methodology.
Is Jack arguing that ID is merely a possibility that future data may (fail to) support? And at what moment do we determine that ID me reliably infered? What data would allow us to reject ID?
Science has a bit of an advantage here of course since natural mechanisms so far have done quite well in explaining nature.
Let me try to give an example which me exemplify why I believe that the ID inference cannot rely purely elimination. Lets assume that I propose the hypothesis that 'XX arose through the help of the pink Unicorn' or 'YY arose through a natural law we have yet to discover'
Now rather than propose testable hypotheses I argue that 'lets accept these hypotheses' on a temporary basis and lets see if other hypotheses can explain XX or YY. Once hypotheses fail consistently to explain XX and/or YY in sufficient detail our hypotheses become 'stronger'.
But do they really become stronger? Is elimination of hypotheses enough to infer that thus 'pink unicorn' or 'some unknown natural law' did it? Or would it be better to assume that we truly have no idea?
If science merely 'scored something as an evolutionary event' when there is no evidence or hypotheses to support this then science also has a problem in that it should have assumed 'we don't know' as the prefered explanation.
An example for instance would be the flagellum, with the present knowledge I do not believe that science can claim that it knows how it DID happen although science surely has proposed ideas how it may have happened. If we use the EF approach this means that until we can eliminate these hypotheses, we should be unable to infer design. Thus there would not be a competing design hypothesis. Now lets assume that ID goes beyond the eliminative nature and actually proposes a hypothesis of design for the flagellum, then we would have various hypotheses all striving to become THE hypothesis. Which means that we have to generate lots of data, predictions, retrodictions and determine which of the available hypotheses works best. Of course that is not a guarantee that the best hypothesis is the correct one but it does help one eliminate hypotheses.
This is why I argue that ID needs to go beyond the Explanatory Filter which relies on elimination but does not propose a competing hypothesis beyond, 'all other hypotheses' cannot explain it.
Does that make sufficient a case for ID? I doubt it.
So perhaps Jack can help us determine how to build a provisional and self correcting methodology for ID. I would argue that such a methodology is already known as the scientific method but I am open to other suggestions. While the scientific method has shown itself to be a very workable and succesful approach, there may be others. After all even the scientific method is tentative and open to improvements.
Until I hear more from the moderator about at least allowing me to respond with more than a single reply per day, I will refrain from additional postings, having reached the customary 3 postings/day limit.
[ 02. November 2002, 17:46: Message edited by: Frances ]
IP: Logged
|
|
Frances
Member
Member # 169
|
posted 03. November 2002 23:15
Teleology and design
Recently on these boards Gina made the following statements which I would like to explore further since I believe that there may be some need to define the term teleology. I also believe that her comments may help us further define predictions, hypotheses relevant to Intelligent Design.
Gina suggest not only that teleology is natural, more natural than other forces, which suggests that teleology can be explored using scientific methods. But Gina also suggest that teleology can be explored empirically through information theory, complex systems and biological emergence.
My first question seems to be a simple one. Gina seems to be suggesting that teleology or 'the design force' is natural, infact more natural than any other since it is metaphysical. Gina also suggests that teleology can be explored empirically through information theory, complex systems and biological emergence. Gina seems to suggest that the design force is purely natural. If that is the case then it seems to me that Intelligent Design cannot be infered purely through the elimination of chance and lawlike regularity. In other words, if teleology can be studied scientifically, what stops us from applying scientific methodologies to teleology?
Gina seems to suggest however that whenever we see adaptation, one sees adaptive problem solving and intelligence in action. This seems to be a remarkable assertion which seems to suggest that 'intelligence' plays an important role in adaptation. Yet, there appear to be countless examples of adaptation which do not seem to require intelligence in action, all that is needed is variation and differential reproduction. Even in cases of 'adaptive mutations' for instance the underlying mechanisms seem to be random variation (random in this case means relative to function) and selection. Don't misunderstand me, this does not mean that intelligence cannot play a role in adaptation. The Baldwin effect for instance shows how such learning can still be quite Darwinian. We also need to determine if the teleology is externally imposed or internal to the organism. Suggestions that problem solving and intelligence play a role suggest to me that Gina is considering a form of internal teleology. Gina's suggestions seem to mirror the direction taken by various researchers, including Shapiro which are looking for genetic engineering toolboxes.
It seems thus quite important that we define exactly what is meant by teleology. Talk.origins has their views on teleology well captured.
Mike Gene has provided us with his insights on teleology and science. Mike argues that teleology is important to science and gives an example of the circulation of blood. What Mike Gene has shown is however not necessarily evidence that teleology contributes to science but rather our use of analogies. Mike seems to be considering the use of such analogies evidence of teleology but we have to ask if this is truely self evident. I would say that perhaps a better explanation can be found in our need for or preference of analogies to think about (new) scientific problems.
Recently Dembski has added to the discussion of teleology
quote:
Design as my colleagues and I are developing it can accommodate the rich contingency and freedom of the natural world and still give scientific content to teleology. To show this, however, will require a mathematical excursion into
evolutionary algorithms and in particular into the No Free Lunch theorems that were proven five years ago.
I wonder if this is what Gina was refering to when she was claiming empirical support to be found in information theory. If so then it would be helpful to explore the extent of the validity of the NFL theorems.
The questions that need to be addressed are
1. What are the No Free Lunch Theorems
2. What are the latest findings wrt these theorems
3. How relevant are the NFL theorems to 'genetic algorithms' as may be envisioned relevant for natural selection?
1. The No Free Lunch theorems refer to work by David Wolpert and William Macready,
Wolpert, D. and W. Macready (1994), "No Free Lunch Theorems for Search"
Wolpert, D. and W. Macready (1997), "No Free Lunch Theorems for Optimization"
They showed that "... all algorithms that search/optimization for an extremum of a cost function perform exactly the same, when averaged over all possible cost functions"
2. Igel and Toussaint have published a paper On Classes of Functions for which No Free Lunch Results Hold (2001. Droste et al have published a paper Perhaps not a free lunch but at least a free appetizer (1998).
Igel and Toussaint show under which circumstance the NFL theorems fail to hold and show that problem classes relevant in practice are not likely to fullfil the requirement for the NFL to apply.
Droste et all argue that quote:
It is often claimed that Evolutionary Algorithms are superior to other optimization techniques, in particular, in situations where not much is known about the objective function to be optimized. In contrast to that Wolpert and Macready (1997) proved that all optimization techniques have the same behavior --- on average over all f : X ! Y where X and Y are finite sets. This result is called No Free Lunch Theorem. Here different scenarios of optimization are presented. It is argued why the scenario on which the No Free Lunch Theorem is based does not model real life optimization. For more realistic scenarios it is argued why optimization techniques differ in their efficiency. For a small example this claim is proved.
These findings suggest that the NFL theorems may not be universal in nature and likely not even applicable to realistic situations. It seems that it is important for the validity of the 'fourth law' that it is determined if indeed NFL theorems apply.
Otherwise we may want to reconsider the following claim by Dembski quote:
Likewise, in biology, even though computational theory is clear that evolutionary algorithms cannot generate complex specified information, by suitably shuffling information around one often gets the impression that evolutionary algorithms can in fact generate complex specified information and that complex specified information is a free lunch after all.
The following link at ARN goes into further detail related to the relevance of NFL to Genetic Algorithms that simulate evolutionary mechanisms. The author argues that if the fitness landscape is allowed to evolve that the NFL theorems are likely not to apply either.
Bruce Gordon very well captures some of my sentiments.
quote:
First of all, what has come to be called 'design theory' is at best a means for mathematically describing, empirically detecting, and then quantifying teleology (goal-directedness) in nature, without prejudging where or whether it will be found. Secondly, if it is granted that teleology might be an objective part of nature, then it also has to be acknowledged that design research can be carried out in a manner that does not violate methodological naturalism as a philosophical constraint on science. I have no attachment one way or the other to methodological naturalism as a metascientific principle, but honesty demands the recognition that design-theoretic research does not logically entail its denial.
Questions that need to be addressed to include:
1. Define teleology. Is teleology internal or external? What do we mean by the term 'Teleology'. Is Teleology as much a part of nature as suggested by Gina? In that case seems self evident that design research can be carried out using the standard methodologies of science.
2. Given that teleology is argued to be empirically detectable, what progress is being made on formulating ways to achieve this using information theory, complex systems and biological emergence? How do the recent findings wrt the NFL theorems affect the empirical detectability of ID?
3. If these forms of teleology are empirically detectable and is found to be part of nature, what does this mean for methodological naturalism?
4. Is teleology a requirement for Intelligent Design? How does teleology apply to the various forms of ID such as interventionism versus front loading?
IP: Logged
|
|
Paul A. Nelson
Member
Member # 26
|
posted 04. November 2002 09:21
Frances wrote:
quote:
Perhaps ID is infered wrt to our present knowledge?
Of course it is. Any scientific inference can be made only with respect to present (not future) knowledge. Why should design inferences be any different, or have to meet a higher epistemological standard?
Frances continued:
quote:
But that would make it vulnerable to become a 'God of the gaps' argument, a placeholder until something better comes around.
But this is the case for any scientific hypothesis. All empirical theories -- design inferences included -- are vulnerable to being overturned in the light of further experience. Design wouldn't be worth much as a predictive theory if it were not vulnerable in this sense.
Many critics of ID seem to have an unreasonable lust for certainty. "We won't infer design unless we can be sure we're not making a mistake." That sort of certainty, however, will not be forthcoming in any empirical investigation.
The unicorn of absolute certainty cannot be captured. But given that the possibility of error is the cost of doing any sort of science at all, I don't see why this should be a worry.
[ 04. November 2002, 10:22: Message edited by: Paul A. Nelson ]
IP: Logged
|
|
Frances
Member
Member # 169
|
posted 04. November 2002 13:04
Paul argues that ID should be infered wrt to our current knowledge. I would agree BUT that would mean that the ID inference will have to drop its requirement/claim that it is reliable in the sense of 'no false positives'.
Indeed all scientific hypotheses are 'placeholders' until something better comes around but ID is somewhat different in that it does not propose a hypothesis of its own but 'merely' claims that at present we do not have regular (law like) or chance hypotheses. My question thus becomes, why do we use ID as a placeholder for our ignorance. What does ID contribute that 'we don't know' does not?
Why should this all be a worry? Simple: Dembski has made the claim that the EF has no false positives. Such a claim is important since the admission of false positives would introduce the possibility of incomplete knowledge and thus the category of 'we don't know'. Incomplete knowledge requires ID to propose a hypothesis itself, not merely infer the absence of other hypotheses.
Perhaps a clarification is needed: Does ID require that there are no false positives in the design inference? Or does ID accept that scientific findings are tentative and that our knowledge is limited? In that case the eliminative filter would seem to be required to become a Bayesian filter in which competing hypotheses go head to head in determining which hypothesis is the best. Perhaps there may be ways to incorporate incomplete knowledge in a Fisherian approach as well?
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 04. November 2002 13:16
Paul Nelson wrote quote:
Many critics of ID seem to have an unreasonable lust for certainty. "We won't infer design unless we can be sure we're not making a mistake." That sort of certainty, however, will not be forthcoming in any empirical investigation.
But it is not ID critics who have evinced that lust for certainty, it is the prime ID theorist, William Dembski. He makes the claim of infallibility of design detections - "no false positives" - for the explanatory filter. See page 24ff of No Free Lunch for a discussion of Dembski's reasons we should believe that the EF is 100% reliable in its detections of design.
RBH
IP: Logged
|
|
Paul A. Nelson
Member
Member # 26
|
posted 04. November 2002 14:14
Frances wrote:
quote:
Paul argues that ID should be infered wrt to our current knowledge. I would agree BUT that would mean that the ID inference will have to drop its requirement/claim that it is reliable in the sense of 'no false positives'.
How does that follow? When Dembski talks about “no false positives,” he is not claiming logical infallibility or certainty for the explanatory filter. Rather, he is making an ordinary inductive generalization:
quote:
In every instance where the complexity-specification criterion attributes design and where the underlying causal story is known...it turns out design actually is present. (NFL, p. 25)
An actual –- not hypothetical –- false positive would show that the complexity-specification criterion atttributed design, but the underlying causal story implicated only natural causes. To my knowledge, no such cases exist.
Dembski notes that design inferences, like all empirical reasoning, can fail:
quote:
...sweeping the field of chance hypotheses is falsifiable in the sense that we might have omitted a crucial chance hypothesis with respect to which an event, though previously exhibiting specified complexity, no longer does so....Archeologists infer that certain chunks of rocks are arrowheads. Detectives infer that certain deaths are deliberate. Cryptographers infer that certain random looking symbol strings are actually encrypted messages. In every case they might be wrong, and further knowledge might reveal a plausible chance process behind what originally appeared to be designed. (NFL, pp. 70-71)
But this is nothing other than the classical problem of induction –- a problem, incidentally, that afflicts all conceptions of probabilistic reasoning. In particular, reconstructing design inferences according to Bayesian methods would not remove the problem of incomplete knowledge. Frances's worries about "incomplete knowledge," therefore, would not be solved by formulating design inferences differently (e.g., as Sober might prefer). Design inferences could still fail in the light of what science might learn in the future, or what we have overlooked in the present.
Frances wrote:
quote:
Why should this all be a worry? Simple: Dembski has made the claim that the EF has no false positives. Such a claim is important since the admission of false positives would introduce the possibility of incomplete knowledge and thus the category of 'we don't know'.
As noted above, the “no false positives” claim is simply an inductive generalization. The claim could be overturned by actual counterexamples; none have succeeded. As for “the possibility of incomplete knowledge,” all design inferences –- like any form of empirical reasoning –- are vulnerable to the problem of induction. Design inferences face no special difficulty here that the rest of science does not.
RBH wrote:
quote:
But it is not ID critics who have evinced that lust for certainty, it is the prime ID theorist, William Dembski. He makes the claim of infallibility of design detections - "no false positives" - for the explanatory filter. See page 24ff of No Free Lunch for a discussion of Dembski's reasons we should believe that the EF is 100% reliable in its detections of design.
As noted above, Dembski is only making a straightforward inductive generalization about the complexity-specification criterion.
[ 04. November 2002, 14:50: Message edited by: Paul A. Nelson ]
IP: Logged
|
|
Genie
Member
Member # 266
|
posted 04. November 2002 14:42
Frances:
quote:
1. Define teleology. Is teleology internal or external? What do we mean by the term 'Teleology'. Is Teleology as much a part of nature as suggested by Gina? In that case seems self evident that design research can be carried out using the standard methodologies of science.
I use Chris Langan's definition for teleology and telic recursion as outlined in his CTMU model. I have many ideas for developing an empirical research program for design and teleology based on his model (as Chris does), and would love to develop them if we should happen to find ourselves in a position to do so. I think this is a very promising future research direction for ID. In the meantime you can look forward to more very interesting design-theoretical work from Chris in future papers.
[ 04. November 2002, 14:43: Message edited by: Gina Lynne LoSasso ]
IP: Logged
|
|
Frances
Member
Member # 169
|
posted 05. November 2002 01:18
Paul raises some good points which still leave us with an important question:
Does the "Design Inference" allow for false positives or not? Dembski seems to be ambivalent about it. On one hand he seems to suggest that it succesfully avoids false hypotheses, on the other hand he seems to accept the risk of false positives as a given when dealing with induction. Yet false positives would severely undermine a design inference based on pure elimination.
Lets first explore why this is an important question. The design inference works on the assumption that by elimination of hypotheses of chance and regularity 'intelligent design' can be infered. But since 'intelligent design' is merely infered through elimination, the possibility of false positives makes the design inference far less useful. Why should this be the case? Well, assume that all known chance and regularity hypotheses can be shown to be rejected, then one might want to argue that Intelligent Design therefor must be true. But what about incomplete knowledge? Unless one wants to argue that we posess omniscience as to all hypotheses, one cannot eliminate incomplete knowledge and thus one cannot eliminate the possibility of false positives.
So what? All inductive arguments after all have this problem. Sure, that indeed is a real problem for inductive arguments but it really is a problem for arguments of elimination. Since we do not posess independent knowledge of the ID hypothesis we are required to eliminate all hypotheses. But given our incomplete knowledge, or given at least the specter of such a possibility, why should we infer intelligent design? Why not infer ignorance? I could make an argument that in all instances in which in the past intelligent design (in biology) was suspected or infered it was shown later that it was merely our ignorance that played tricks with us.
One may ask, why does he places biology in brackets? Well, Dembski, in his argument to show that there are no false positives argues that in all instances in which the EF was used, design was reliably infered. But that of course is begging the question. At most Dembski can show a few examples in which the competing hypotheses were intelligent design or chance. The examples were simplistic and not really relevant to situations in which we are attempting to detect intelligent design in nature.
Wilkins and Elsberry refer to such design as rarefied design, no inductive knowledge exists about this design. One may attempt to argue from analogy but such approaches will have the same problems that were the downfall of Paley's arguments.
False positives are a real problem for elimination because the only evidence for the hypothesis is the absence of evidence. But if the absence of evidence could be due to our ignorance then we should not jump to conclusions.
So yes, all inductive arguments have this problem. However the design inference has two additional problems, its detection depends ultimately on the accuracy of rejecting chance and regularity hypotheses. One may wish to tentatively infer design but why not tentatively infer ignorance until we have other means to test our ID hypothesis? Why should ID be infered under such circumstances? I believe no good reason can be given for the special position of the ID hypothesis. Secondly, inductive arguments depend strongly on the validity of the subset to be extended to the common. Given the paucity of samples in which the EF was tested, making more general claims would seem to be a fallacy of hasty generalization.
In a future edit I will address the issues raised by Sober about the fact that no probabilistic equivalent of modus tollens exists.
Paul states that
quote:
An actual –- not hypothetical –- false positive would show that the complexity-specification criterion atttributed design, but the underlying causal story implicated only natural causes. To my knowledge, no such cases exist.
Perhaps Paul is right but perhaps he isn't. And why do these false positives have to be actual not hypothetical? After all there are only a few cases in which the EF has been applied and none seem to have much relevance. I could envision various scenarios in which incomplete knowledge would lead to false positives. Since one cannot argue that we have complete knowledge, the potential for false positives exists. Even if Paul may not be aware of such cases, there is no real way to determine if he is correct. In fact I would argue that good reasons exist to suspect that such scenarios do exist, although that does not mean that Paul should be aware of them.
And as I have shown the design inference has a significant problem compared to the common scientific approach namely that as formulated the design inference is nothing more than 'chance and regularity cannot explain it' and although we realize that there is the possibility that our knowledge is incomplete we will infer design.
And what then? Why not infer 'we don't know' until we can formulate an ID hypothesis? Why should the ID inference be given a priveleged status among 'hypotheses' in that it does not require any hypothesis? What makes the ID different from 'we don't know' which fares equally well through an absence of hypothesis for our ignorance?
Paul is also correct that this problem affects all probabilistic reasoning, but the EF due to its privileged status seems very susceptible to this problem since incomplete knowledge seems inherent and thus the EF will always remain tentative, as tentative as the hypothesis 'we don't know'.
There are additional problems with hypotheses that are not logical complements but probabilistic complements. As Sober has shown, they are not. In fact they are probabilistic arguments and there is no probabilistic equivalent to the modus tollens argument.
So how can we liberate the ID inference from its ties to 'we don't know'? I would argue by proposing a testable ID hypothesis which can run side by side with other scientific hypotheses, and while a Bayesian approach will not help us address if a hypothesis is correct, we can use it to determine which hypothesis is the best one. No such ability exists for the Fisherian approach since it does not require an ID hypothesis to be formulated.
So
Dembski in "Mere Creation" suggests the following
quote:
... I want to argue that the explanatory filter successfully avoids of false positives. Thus whenever the explanatory filter attributes design, it does so correctly. Let us now see why this is the case. I offer two arguments. The first is a straightforward inductive argument: in every instance when the explanatory filter attributes design and where the underlying causal story is known, it turns out design is present; therefore design actually is present whenever the explanatory filter attributes design.
Dembski (MC) Page 107
Source
We should in fact realize that 'succesfully' is limited to a few basic examples and that the formulation of this inductive claim is at least hasty and likely erroneous. The EF will have false positives, how to deal with in a meaningful manner requires a formulation of an ID hypothesis, or we have condemned the ID inference to be identical to the 'we don't know' inference. And likely Ockham's razor would have some shredding to do.
===========Update=============
A good example where claims of 'no false positives' may lead is the response to Dembski about the Oklo natural reactor.
quote:
But suppose the Oklo reactor ended up satisfying the criterion [complexity-specification]. Would this vitiate the complexity-specification criterion? Not at all. At worst it would indicate that certain naturally occurring events or objects that we initially expected to involve no design actually do involve design.
Seems like CSI as formulated by Dembski would be unfalsifiable? Of course it has to be since allowing for false positives would undermine the whole filter. And yet there seem to be various reasons to believe that indeed false positives exist and that CSI may not be a reliable indicator of design after all.
Thus my suggestion to rather than eliminate, ID proposes its own ID hypotheses to be tested side by side with other hypotheses. After all we know that such approaches do work quite well in areas such as criminology, archaeology etc.
This also brings me to the following interesting argument by Murray: Intelligent design comes in two flavors, 'intervention' and 'front loading' but for all practical purposes front loading would mean that the methodological naturalistic approach of science would be sufficient and if we cannot distinguish if ID events are intervention or front loaded, why should we accept that ID proposes an alternative to methodological naturalism? It seems to almost embrace it. But this would seem to contradict with the EF approach in that ID is Not (chance and regularity) but if it is Not(chance and regularity) then it seems to be extra-natural.
quote:
However, this seems to entail a claim that many IDT advocates deny, sometimes strenuously, namely, that appeal to intelligent design requires appeal to supernatural intervention in the course of nature. One cannot have it both ways. If my success at explaining an event’s occurrence through law-like processes undercuts a design explanation, then the presence of design requires that some events be caused in a non-nomically regular way, i.e., miraculously. This is simple modus tollens.7
And finally I promised to return to the 'modus tollens' fallacy. While modus tollens arguments such as if X then Y, and we find not Y then we can conclude not X. But the fallacy is that this argument is repeated for probabilistic arguments.
quote:
This paper defends two theses about probabilistic reasoning. First, although modus
ponens has a probabilistic analog, modus tollens does not – the fact that a hypothesis says that
an observation is very improbable does not entail that the hypothesis is improbable.
Source
Thus probabilistic arguments of the form modus tollens seem to be illegitimate.
Fitelson has some good remarks on the philosophy of statistics in which he raises some of the same issues (page 6).
There seem to be a lot of areas of confusion, or contention wrt the Design Inference and it will be helpful to categorize them and explore how we can strengthen the ID design-theory. Given the myriad of problems with the EF it seems almost inevitable that ID rather than propose elimination of all alternatives proposes its own design hypotheses. In my 'the future of ID' I have given some examples of such predictions, explanations that may help formulate a non ad-hoc design hypothesis.
Once such a hypothesis exists we can start exploring how it compares to its competitors.
[ 06. November 2002, 00:31: Message edited by: Frances ]
IP: Logged
|
|
Terence A-H Tan
Member
Member # 234
|
posted 05. November 2002 08:14
Another meager 2 cents worth of thought from my spare time mullings.
It seems intelligent design has come to mean different things for different individuals constructing different theories from different epistemilogical backgrounds. Hence what seems to be intersecting points of contentions would benefit from some clarification, otherwise all debate comes to nought if we're merely interested in looking at our own pages.
In my mullings, I was wondering, what if we turn the EF the other way round ending with random chance. What makes random chance easier to prove than intelligent design ? One reason is that random chance is something that continues to exist. We can invoke it by the toss of a coin and prove its mechanism. Intelligent design, in its original proposed form however belongs to a different category. Unlike random chance, it is historical or archaeological, albeit it seems some may have liken it to random chance in the sense that it continues to exist. The uniqueness of being historical is that what is done by an intelligent agent has been done. The work is finished. Unless the intelligent agent continues the work at present time and is observable and thus testable, it is more difficult for intelligent design in its original conception to propose confirmatory scientific research because it is dealing with something archaelogical. Hence for this kind of historical science, falsifiability/elimination is more useful. One must not confuse the EF used for specified complexity as equivalent to intelligent design. Specified complexity is only 1 of 2 confirmatory scientific research that supports the notion of intelligent design. The other being irreducible complexity. What Dembski may be seeking is adding more pillars (other types of confirmatory scientific research) to these original 2 to support the notion of intelligent design. This requires creative leaps of insight.
To end this mulling, i can see where other variants of intelligent design is coming from and their research interests may thus differ.
IP: Logged
|
|
Frances
Member
Member # 169
|
posted 06. November 2002 00:12
CTMU 'a theory of everuthing'?
I have read the various web pages describing CTMU and I would like to ask a few questions. I apologize for any questions which may seem self evident.
I have seen a lot of claims of what the CTMU is claimed to be able to achieve but I have seen few examples of this being actually the case. If the CTMU is presented as _a_ 'Theory Of Everything' then my first question would be if it can explain everything, what does it really explain?
For instance abiogenesis was mentioned when discussing a theoretical self-deterministic and self-creative design 'force'. Gina argues that abiogenesis is not even a well formed hypothesis since hypotheses 'require models' but that seems to be making some assumptions, namely that hypotheses require models. What do you mean my models? Mathematical models? Logical constructs? Abiogenesis surely seem to have the requirements for such, the question is does CTMU provide us with such a model? What does CTMU's model explain as far as abiogenesis is concerned? What does it predict? How can CTMU be put to the test in other words?
Abiogenesis proposes for instance the RNA world, it analyzes known DNA sequences to determine the relative abundance of amino acids at the time of formation similarly Molecular evolution before the origin of species..
How does the TOE fit in with these data? Are these data expected by this theory or at odds with this theory?
All that abiogenesis hypotheses do is use known models of chemical reactions, combined with various assumptions to make predictions. So the suggestion that abiogenesis does not even have a preliminary model seems to be begging the question. Just because molecules and reactions require complex laws of nature should not disqualify abiogenesis from considerations? In fact it is exactly the nature of these reactions that make for good abiogenesis candidates. And actually many of these complex reactions can be quite easily studied. Do we need to understand every detail of these reactions? Of course not. We need to be able to apply our acquired knowledge as best as possible to propose (various) hypotheses of abiogenesis.
If we are waiting until we understand everything then there is really no need to explain it anymore nor is there any reason to do science but without science how do we acquire the knowledge to 'know everything'?
Thus hypotheses which use the best knowledge and data to propose a possible explanation for the observed facts and make predictions of non-observed facts. What does CTMU do? It makes a lot of claims about self-selection, tautologies, self-aware universes, mechaanisms for intelligent design. In fact I wonder what CTMU does not explain? But more importantly what does CTMY contribute to science? Has it been able to explain the missing solar neutrinos? Does it propose new laws of physics? Does it introduce new knowledge? The proof would be in 'the pudding' so to speak and I have yet to see evidence of such pudding. If it is a theory of everything, it surely explains evolution, abiogenesis and many other features for which we may not know all the details. So what were we missing in evolution and/or abiogenesis? What mechanisms were lacking? How can we improve our theories now that we have a theory of everything?
Additionally I am under the impression that requiremens such as self-awareness are unnecessary and ad hoc. They do help introducing telic principles but lets assume the following:
Our universe and reality is totally explained from the initial and boundary conditions and natural laws. In fact, thsee boundary conditions which are for all practical purpose hidden from us behind a one-way mirror known as the Planck time, set in motion a self generating, self perpetuating universe. With one difference, it has no self-awareness, no goals. In fact one may wonder why a self aware, self generating universe seems to have chosen to select a path which leads to its own inevitable demise, 'heat death', the second law of thermodynamics seems to be more powerful than the universe. Are laws somehow untouchable to our self aware, self generating universe? But if that is the case what explains these laws? A tautological argument based on arguably consistent logic makes for a nice thought experiment but what practical applications does it have? And I am not asking what practical applications it may have, we can all theorize about it, what I would like to see is some evidence that this TOE has any practical purpose.
What if the Universe were one big non-linear partial differential equation with given initial and boundary conditions? No need for self-awareness. Seems like an acceptable reality to me and quite a workable one as well.
I would be interested to hear more about how CTMU's ideas fit in with ID. Seems that CTMU has no need for an external intelligence and that it considers everything explainable by natural law? Am I correct in this?
Langam's teleology seems to be at odds with present ID concepts and theories? How do you suggest can the ID inference be improven with the use of the 'theory of CTMU'? What research directions do you foresee that will contribute to design-theoretical research? And how does CTMU fit it with design theoretical research at this moment?
IP: Logged
|
|
Genie
Member
Member # 266
|
posted 06. November 2002 10:41
quote:
I would be interested to hear more about how CTMU's ideas fit in with ID.
See the last section of his paper ("The CTMU and ID"). That's one of the areas Chris is working on. He may even be writing a paper addressing ID more specifically, but I can't speak for him. I do know that he is working on developing different aspects of the model. We try to make new material available as it is developed. Chris also holds focused discussions from time to time. He just finished a month-long discussion that was open to ISCID members (the archive is still there and Chris will be adding a couple of responses to late questions.) When he has time, he'll host another, but probably something shorter, maybe over a weekend.
IP: Logged
|
|
|
Printer-friendly view of this topic