|
Author
|
Topic: PIPR: Principle of Inflated Probabilistic Resources
|
William A. Dembski
Member
Member # 7
|
posted 25. February 2002 08:19
One often hears these days that a purely naturalistic origin of life is quite likely, perhaps not in the observable universe, but when all the other universes that exist beyond our universe are factored in (inflationary cosmology and quantum many worlds are often invoked here). Against such claims, I want to propose the following principle, which we may dub the Principle of Inflated Probabilistic Resources: "If a theory requires vastly more stuff than is available in the observable universe to render it plausible (exceedingly many probabilistic resources is how I typically put it in my writings), then it is really not an insightful scientific explanation and we need to keep looking." The following considerations justify this principle, at least in my mind. If there are infinitely many universes similar to ours, then there are universes in which tractors and automobiles and microwaves assemble themselves as vastly improbable thermodynamic accidents. To explain such artifacts as the result of chance is singularly unhelpful because in the vast majority of cases such artifacts will be the result of design (i.e., on most planets within an inflationary universe such artifacts won't result from chance but from design). Thus even if natural selection and random variation on an odd planet in an inflationary universe could by some vast improbability bring about some biological system, that would not show that natural selection and random variation rather than design was responsible for such a system on our planet because the Darwinian mechanism may be largely ineffective at bringing about such systems and design, on the other hand, might be perfectly suited to that task (I have in the back of my mind here Michael Behe's irreducibly complex molecular machines). Design is not an explanation of last resort. Just because chance could, as some bare possibility or "frozen accident," do something does not mean that we need give it a second thought unless it renders it reasonably probable in our corner of the universe. I deal with this line of reasoning at greater length in my paper "The Chance of the Gaps" posted in the ISCID archives, though I don't there explicitly formulate the PIPR.
IP: Logged
|
|
|
|
William A. Dembski
Member
Member # 7
|
posted 25. February 2002 14:21
In reply to Jeremy: ID commits one to a generic intelligence, not to an supernatural intelligence. So it's not at all clear in what sense ID commits one to "vastly more stuff." And even if the intelligence ends up being unembodied or supernatural, the intelligence would be of a fundamentally different kind from physical stuff, so again there is no question of quantifying the stuff that is supposed to render a vastly improbable event probable. The bottom line may just end up that you've got to choose your metaphysics: a world in which unembodied designer(s) are real or a world in which the amount of physical stuff vastly exceeds anything to which we have empirical access (or perhaps some other options). I don't think it's an accident that the physics of inflationary cosmology and quantum many worlds is blurring the boundaries with metaphysics.
IP: Logged
|
|
Bryan Cross
Member
Member # 51
|
posted 25. February 2002 15:15
Even if ID only commits one to a generic intelligence, and therefore does not require 'lots more stuff', positing 'lots more stuff' is not the only way to inflate probabilistic resources, since one could also inflate probabilistic resources by positing more 'kinds of things'. To define the principle this way seems to beg the question, favoring qualitative expansion over quantitative expansion. For any abductive problem, one or the other type of expansion may be necessary. Therefore, we can't simply rule out positing more stuff than we have directly observed. For any abductive situation, it would seem, the proper method is to compare the posited 'more stuff' explanation with the 'more kinds' explanation, and see which is superior.
IP: Logged
|
|
Jeremy Alder
Member
Member # 32
|
posted 25. February 2002 17:39
quote:
Originally posted by Bryan Cross:
Even if ID only commits one to a generic intelligence, and therefore does not require 'lots more stuff', positing 'lots more stuff' is not the only way to inflate probabilistic resources, since one could also inflate probabilistic resources by positing more 'kinds of things'. To define the principle this way seems to beg the question, favoring qualitative expansion over quantitative expansion. For any abductive problem, one or the other type of expansion may be necessary. Therefore, we can't simply rule out positing more stuff than we have directly observed. For any abductive situation, it would seem, the proper method is to compare the posited 'more stuff' explanation with the 'more kinds' explanation, and see which is superior.
I'm going to agree with Bryan here. It seems to me that we are expanding probabilistic resources by positing a different kind of thing, a designer. Not only that, but whether or not the designer is ultimately supernatural, it is an appeal to something outside of the observed universe given that we don't have any independent evidence for any plausible embodied designer for the OOL inside the observed universe. Of course, there very well may be such ETI, but at this time we have no independent evidence for them. It seems to me that if one is going to require independent evidence or reasons for multiple universes before considering it as an "explanation" for something, as in 2.8 of NFL, then it seems legitimate for the ID critic to require independent evidence of the posited designer(s)before considering it as an explanation. Both are expansions of probabilistic resources that exceed the emperical evidence. Personally, I don't think one needs independent evidence of the designer before being justified in using ID as a working hypothesis, so I'd be inclined to discard the principle if the situation I outlined above follows from it. [ 25 February 2002: Message edited by: Jeremy S. Alder ]
IP: Logged
|
|
William A. Dembski
Member
Member # 7
|
posted 25. February 2002 17:50
At issue here is what has the causal power to produce things that exhibit specified complexity, irreducible complexity, functional complexity, call it what you will. Intelligence does, and we know that intelligence exists. So we are not positing anything fundamentally new or strange. On the other hand, invoking chance to explain objects with these features works only if one is willing to countenance events of high improbability where without substantial probabilistic resources there would be no reason to assign any plausibility to them. For these reasons, I would say that we can have independent evidence for a designer whereas we don't and cannot in principle have independent evidence for unlimited probabilistic resources.
IP: Logged
|
|
|
|
Columbo
Member
Member # 113
|
posted 25. February 2002 17:59
Oops! That's hieroglyphs. (Pardon me!)
IP: Logged
|
|
Tom Stalnaker
Member
Member # 114
|
posted 25. February 2002 19:24
I just want to point out a difference between using inflationary probability to explain life naturistically, and using it to explain how human-designed objects could be explained without design. The former hinges on the logic that, out of all the universes that putatively exist, we can logically only be present in one that contains life. This would explain why, in the context of a multitude of universes, we are present in one of the almost infinitely small subset of universes in which life evolved by chance. When Dembski extends this logic to human-designed objects (in order to show how inflationary probability would destroy all design inferences), he does not acknowledge that the logic described above no longer applies. That is, we could be present in either a universe in which tractors are all designed OR one in which tractors are all randomly self-assembled. But it is vastly more likely that we are present in the former kind of universe, no matter how many universes there are overall.
In an inflationary multitude of real universes, therefore, a very different logic can be used with life and non-life. It is logically required that we are present in one of the subset of universes in which life exists. It is not logically required, however, that we be present in one of the sub-subset of universes in which life exists and also in which all of what we think of as human-designed objects are in fact not designed. Therefore, if those inflationary probabilities are true, we can still use the design inference on non-life, but we cannot use it on life.
By the way, I am not saying I buy the inflationary probability argument, and I am sympathetic to the design inference about life, but it seems to me that Dembski's criticism of that argument (as described in the paper and the post)is not quite right.
IP: Logged
|
|
|
|
|
|
William A. Dembski
Member
Member # 7
|
posted 26. February 2002 10:03
Tom Stalnaker claims that the PIPR stumbles on a selection effect. The selection effect argument for trying to get around small probabilities has been thoroughly debunked now for about 15-20 years (Richard Swinburne, Francis Crick, John Leslie, and Bill Craig among others have debunked it). The key problem with it is that it confuses a necessary condition for an explanation -- just because it's necessary for us to be here to observe some feature of the universe doesn't explain that feature. Consequently, chance works no better in accounting for human observers than in accounting for randomly formed tractors. I cite that literature in _The Design Inference_ and advert to the problem in _No Free Lunch_. Have a look.
IP: Logged
|
|
Tom Stalnaker
Member
Member # 114
|
posted 26. February 2002 18:18
Dembski says that "just because it's necessary for us to be here to observe some feature of the universe doesn't explain that feature". However, no one invokes the selection effect to explain the feature (i.e. intelligent life) itself; rather, one would invoke it only to explain why we happen to be in the extremely rare place that it did occur. The former is explained (in this argument) by the very large probabilistic resources of the expansionist universe.
Let me just briefly describe a thought experiment, which is an altered version of the one Swinburne used to debunk the "anthropic principle" (aside: doesn't Steven Hawking still espouse the anthropic principle?): Suppose that 10^19 madmen kidnap 10^19 other men and put them in 10^19 rooms with card-shuffling machines that will kill the man in that room unless it draws an ace of hearts from each of 10 decks. Each man will only see the cards in the unlikely case that aces of hearts are drawn from all the decks. All of the men are told that there are 10^19 other men in similar situations. If a man then finds himself looking at 10 aces of hearts, what should he conclude? I think he should conclude (assuming he has no other reason to doubt what he was told) that he is one of the lucky few whose machine randomly drew 10 aces of hearts. This is an illustration of the selection effect, and it will only work, of course, if there are sufficient replications to overcome the low probability of any one specified event happening. But that is exactly what the expansionist universe postulates (not a multitude of possible universes, but a multitude of real sub-universes).
Let me just change the experiment slightly to illustrate another case: Suppose that everything is the same as above, except that instead of killing the victim if the machine fails to draw 10 aces of hearts, the machine will simply drop a mouse onto his lap if it draws 10 aces of hearts, and do nothing otherwise. A particular man finds himself with a mouse on his lap, and what should he conclude? He should conclude that the machine is not working as advertised. Because there is no selection effect in this case, he comes to a different conclusion than in the first thought experiment. There will probably be some men with mouses on their laps, but one of them is not likely to be the particular man that we are talking about. The second experiment illustrates, in my mind (and please point out where I am wrong), why the design inference would still work for objects, even in an expansionist universe.
IP: Logged
|
|
Petri Kivenheimo
Member
Member # 151
|
posted 28. February 2002 16:11
This is some musings concerning the selection effect idea.
First, a try to rephrase the idea itself (mainly in order to let the enlightened
reader correct misunderstandings):
Any (somewhat stochastic?) cause to the effect of mass extinction among some population will render the survival of the survived a natural thing in no need of further explanations under the assumption that the pre-extinction phase population was numerous enough to give a plausible probability to the case that some of its members might survive the disaster.
Now, a question:
The rationale behind this idea seems to be simply that of a lottery winner: given enough participants, it is probable (and according to the rules of the lottery, perhaps even sure) that somebody will come out as a winner. The identity of the winner is now explainable by chance alone, if there are no further facts to be taken under consideration.
This rationale is however equally applicable to the case in which the other participants did not pass away (as in lottery generally is the case). So the last reasonings in Tom Stalnaker's message seem somewhat surprising to me. That could be because the proposed rationale behind the survival effect idea were something else than the above one, and I'd be interested to hear about what it could be.
(The very title "selection effect" could perhaps be taken as a hint that the idea were not grounded on a pure chance after all, but on some kind of "survival of the fittest". In that case, a more appropriate comparison than that of a lottery winner might be that of an athletics winner: often the result of a competition was foreseeable to the commentators anyhow. In this case, though, the thought experiments with the card decks would seem out of place to me.)
IP: Logged
|
|
Tom Stalnaker
Member
Member # 114
|
posted 28. February 2002 19:26
I could be totally wrong on what I said earlier, in which case I would be especially  given that I have diverted the discussion of the board. In that case accept my apologies.
That being said, I still think the second case, in which there is no "selection effect" (I am using that not in any evolutionary way) is somewhat different than the case in which survival means that you won the lottery. For instance, imagine a lottery in which there are 10^19 players, and in which each player does not have any contact with other lottery players, and in which no one tells you if you have won. If, after the lottery is over, you are considering whether you won the lottery but have no independent evidence of that being the case, you should conclude that you probably did not win it. (this is a bit different than my mouse example, but more appropriate, I think)
This kind of lottery is analagous to Dembski's Arthur Rubenstein case, in which we are wondering if Arthur Rubenstein (whom we are listening to) might be a musical imbecile who is very lucky while banging at the keyboard. Just as in the lottery, no one would "tell" us if Rubenstein were a lucky musical imbecile (i.e. we have no independent evidence of it.) Well, suppose we are living in an expansionist universe. Then it is likely that somewhere in the expansionist universe, someone resembling Rubenstein is in fact a lucky musical imbecile. However, it is extremely unlikely to be the case in our little corner of the universe. Thus we are entitled to conclude that Rubenstein is not a lucky musical imbecile. Thus the design inference.
This kind of reasoning is quite different in the case in which all non-winners have already been killed. In that case, our existence gives us a piece of information that we don't have in the other case. This piece of information, as far as I can tell, is valid and can be used to inform our probabilistic assessment. This information is analagous to that of being given 100,000,000 dollars after the lottery has been conducted. In that case, we would be surprised, but we would be in possession of independent information indicating that we in fact won the lottery. (When I say independent information, I mean information independent of our pre-lottery calculation of the probability of winning).
So, the two kinds of cases appear to elicit different post-event analyses. In the case with a selection effect, we are surprised and feel extremely lucky, but still attribute the result to chance, whereas in the case with no selection effect, we strongly doubt whether the small chance event even happened. (it occurs to me that I may be misusing "selection effect" , but I hope you know what I mean. I was trying to imitate Dembski's usage  ).
IP: Logged
|
|
|
Printer-friendly view of this topic