Engineering Rationality: Philosophy Forums

Philosophy Forums » General Philosophy » Engineering Rationality » Engineering Rationality

Engineering Rationality
As compared to scientific rationality

Page: 1 2

Engineering Rationality

aletheist
Professor

Usergroup: Sponsors
Joined: Feb 06, 2009
Location: Olathe, KS

Total Topics: 12
Total Posts: 623

Posted 06/30/09 - 05:50 PM: Subject: Engineering Rationality	#1
This is my third thread on the philosophy of engineering, and again I look forward to your feedback. In the previous ones, I discussed the differences between the scientific method and the engineering method, and between scientific knowledge and engineering knowledge. This time, I would like to explore the related subject of the types of reasoning that scientists and engineers typically employ, and why this distinction is important from a philosophical perspective. My thoughts on this subject have been shaped by a series of papers written over the last twenty-five years by Dr. Steven L. Goldman, the Andrew W. Mellon Distinguished Professor of the Humanities at Lehigh University in Bethlehem, Pennsylvania. The most recent one is, "Why We Need a Philosophy of Engineering: A Work in Progress" (Interdisciplinary Science Reviews, Vol. 29, No. 2, June 2004, pp. 163-176). Dr. Goldman points out that scientific reasoning is primarily concerned with the concepts of necessity, certainty, universality, abstractness, and theory. It seeks objective knowledge of timeless truth that is based on reality, for the purpose of intellectual contemplation and understanding. By contrast, the reasoning of engineers is characterized by contingency, probability, particularity, concreteness, and practice. They rely on subjective beliefs and historical opinions that are derived from experience, with the goal of willful action and use. As an illustration, consider the example of a bridge. There is no single optimal span for a particular location, although a solid case can be made that the one across the Golden Gate comes pretty close. A staggering array of variables contributes to establishing the type, alignment, materials, height, etc. Tradeoffs are inevitable because of legal restrictions, budgetary constraints, and many other considerations, only some of which are explicit, and many of which are not even technical. In the end, it is the collective (and fallible) judgment of the design team that dictates the final form of the structure, rather than a rigid (and inerrant) formula. Dr. Goldman presents the approaches of science and engineering under two headings: the Principle of Sufficient Reason (PSR) and the Principle of Insufficient Reason (PIR), respectively. Strictly speaking (per Wikipedia), PSR states that anything that happens does so for a definite reason, while PIR - also known as the Principle of Indifference - states that if there are multiple mutually exclusive and collectively exhaustive possibilities that are indistinguishable except for their names, then each one should be assigned the same likelihood. PSR essentially assumes that there is always one "right" solution to every problem, while PIR requires an intentional choice from among several equally valid alternatives. Dr. Goldman goes on to suggest that Western culture has long favored PSR over PIR, and that this bias has contributed to the generally high regard in which science is held (relative to engineering) to this very day. In fact, the conflict within the discipline of philosophy goes all the way back to Plato’s harsh criticism of the Sophists, the ancient Greek champions of rhetoric. His perceived triumph in that exchange is reflected in the negative connotations that words like "sophistry" and "rhetorical" still carry some twenty-four centuries later. Plato relentlessly contended that the Sophists were only interested in teaching tricks for winning arguments and were ignorant of the good, the right, and the true. He sought the ideals of pure reason and perfect justice, which the Sophists rejected as unrealistic; instead, they advocated social discourse and pragmatic action. The original function of classical rhetoric was to teach citizens in a democracy how to make and justify seemingly arbitrary decisions in a context of uncertainty - exactly the task of modern engineers, as well as free human beings throughout the ages. Ironically, scientists would put themselves out of business if they could ever actually achieve full comprehension of the mysteries of nature. Karl Popper insisted that only propositions that are "falsifiable" - capable of being disproved - should ever be described as "scientific". Thomas Kuhn popularized the idea that science only advances significantly when one paradigm is replaced by another, usually because it fits the data better. Theories are not discovered; they are selected from a number of plausible explanations, then tested and modified as necessary, and are always subject to being revised or discarded. In other words, while drawing contrasts between scientific and engineering reasoning has been common historically, it really sets up a false dichotomy, as Dr. Goldman observed; in the absence of complete knowledge, PIR is the only feasible option. It is precisely when there is more than one path available to follow that it is possible and desirable to exercise wisdom - sophia in Greek, from which the Sophists took their name. Consequently, because of their training and temperament, engineers should be uniquely suited to help society wrestle with the many challenges that it faces - not just in the technological realm, but in all areas of life.
"Be attentive, Be intelligent, Be reasonable, Be responsible." - Bernard Lonergan (1904-1984)

Incision
Professor

Usergroup: Administrators
Joined: Jan 04, 2008
Location: Utah

Total Topics: 23
Total Posts: 906

Posted 07/01/09 - 09:07 AM:	#2
I also think Plato gave the Sophists too bad a rap, but I'm not too hot on the PIR. I'm having a little troubled reconstructing your argument; would (part of) it go like this? (1) For any set of exhaustive and mutually exclusive alternatives, the probability of each alternative is equal. (PIR) (2) Therefore, if multiple alternatives have the same probability, it is epistemically permissible to choose one arbitrarily. (Also PIR?) There may be some confusion over what the PIR is: I'll assume that it's (1), and that (2) is an inference from it. As for this argument then, (1) is false. There are many ways of "dividing up" alternatives, so the PIR gives inconsistent results. Here's an exhaustive and mutually exclusive set of possibilities: {A, ~A} Applying the PIR, we conclude that P(A) = 1/2. Here's another: {(A & B), (A & ~B), ~A} Applying the PIR, we conclude that P(A) = 2/3. Here's another: {(A & B & C), (A & B & ~C), (A & ~B), ~A} Applying the PIR. . . . We can give P(A) any value just by changing our description of the set (if we want to lower P(A), we just repeat the same trick with ~A). There may be some resistance to this counterargument; obviously, the last set is a silly set, while the first is natural and intuitive. So maybe we could preserve the PIR just by adding a clause to the effect that the set must not be silly. But here's an example from the SEP. A machine at regular intervals shoots out toy blocks, of a random size between 1 meter on a side and 0 meters (in which case, I suppose, it shoots out nothing). What's the probability that it shoots out a block somewhere between 0 meters and 1/2 a meter on a side? The non-silly answer seems to be 1/2. Now, what's the probability that it shoots out a block somewhere between 0 m² and 1/4 m² on a face? The non-silly answer seems to be 1/4. But notice that blocks 1/2 m on a side and blocks 1/4 m² on a face are the same size. So we get contradictions even when we're being earnest. One last example: what's the chance of rain? If an earnest set is {rain, ~rain}, then P(rain) = 1/2. But here in Utah, it's a desert: P(rain) should be low. So the set the gives the right answer is something like the silly {rain, (~rain & A & B), (~rain & A & ~B), (~rain & ~A)}. The PIR gets its plausibility because it does sometimes give just the right answer, when used with good sense and in certain contexts like gambling. But how do we know to make the set {heads, tails} and not the set {heads, tails, edge}? I'd suggest it's because we never see coins landing on edges -- nothing to do with math. So even used with good sense and treated as a heuristic, the PIR would give the wrong answer any time we're not sure what to include in the set, or the probability of each member of the most intuitive set is unequal. But even were the PIR true, (2) would not follow. As a minimum standard for epistemic permissibility, you should not know that the proposition is false, or very likely false. But for any large set of equally-likely outcomes, the probability of any particular outcome will be small. Well, that doesn't touch on everything in the OP, but maybe it's enough for one post.

mway
Professor

Usergroup: Members
Joined: Apr 07, 2009

Total Topics: 14
Total Posts: 509

Posted 07/01/09 - 03:17 PM:	#3
I think you missed the important word 'exhaustive'. To address the argument, you really need to apply real world situations that aren't reducable to some small definable set. You could use the bridge example above, or for this post, I will use software engineering. After reviewing the specification for any piece of software there is an arbitrarily large number of ways to achieve the result, whilst also being equally correct. You are right in saying that certain outcomes will most likely have different probabilities, but the probabilities cannot be calculated, so some arbitrary choice must be made. This can be applied to all engineering endeavors.
Lame is to Wav, as the Brain is to Reality.

aletheist
Professor

Usergroup: Sponsors
Joined: Feb 06, 2009
Location: Olathe, KS

Total Topics: 12
Total Posts: 623

Posted 07/05/09 - 05:42 PM:	#4
Incision wrote: (1) For any set of exhaustive and mutually exclusive alternatives, the probability of each alternative is equal. (PIR) You seem to have left out a critical characteristic of the alternatives--for PIR (strictly speaking) to apply, they must not only be exhaustive and mutually exclusive, but also indistinguishable except for their names. Also, an engineer is not usually concerned with "exact" probabilities; he or she needs to make decisions that are "good enough" in a context of (usually unquantifiable) uncertainty without exhausting the available resources. mway wrote: After reviewing the specification for any piece of software there is an arbitrarily large number of ways to achieve the result, whilst also being equally correct. You are right in saying that certain outcomes will most likely have different probabilities, but the probabilities cannot be calculated, so some arbitrary choice must be made. This can be applied to all engineering endeavors. This was really my main point, more so than PIR per se. Some choices are "better" than others, depending on what evaluation criteria you use and what tradeoffs you accept as part of the "optimization" process; but no single choice is (objectively) the "best" option.
"Be attentive, Be intelligent, Be reasonable, Be responsible." - Bernard Lonergan (1904-1984)

Incision
Professor

Usergroup: Administrators
Joined: Jan 04, 2008
Location: Utah

Total Topics: 23
Total Posts: 906

Posted 07/05/09 - 10:35 PM:	#5
mway wrote: I think you missed the important word 'exhaustive'. aletheist wrote: You seem to have left out a critical characteristic of the alternatives--for PIR (strictly speaking) to apply, they must not only be exhaustive and mutually exclusive, but also indistinguishable except for their names. But the sets {A, ~A} {(A & B), (A & ~B), ~A} are exhaustive and constituted of members indistinguishable except for their names. They are exhaustive in virtue of the law of excluded middles: a set of propositions is exhaustive just in case on member must be true, and obviously of those sets one member must be true. They are indistinguishable except for their names because they listed simply as "A," "B," etc.: they are variables ranging over all propositions. Isn't that right? Do you see a feature A has that B doesn't?

Incision
Professor

Usergroup: Administrators
Joined: Jan 04, 2008
Location: Utah

Total Topics: 23
Total Posts: 906

Posted 07/05/09 - 10:49 PM:	#6
If it makes you feel any better, the PSR is also incoherent (where is PSR is that there is a good explanation for any proposition), so if the engineering and scientific models have respective commitments to the PIR and PSR, then the engineering model is no worse on that account. Let P be the conjunction of all contingently true propositions. Then P itself must be a contingently true proposition. So if the PSR holds, then there should be a true proposition E which explains P. This E is either contingently true or necessarily true. But it can't be necessarily true, since if it entails P, and P is contingent, then P is necessary, which by hypothesis is false. So E must be contingently true. But if so, then E is a part of what it is supposed to explain, so it would be a circular explanation. But it's impossible for a good explanation to be circular, so by reductio we must reject the PSR.

aletheist
Professor

Usergroup: Sponsors
Joined: Feb 06, 2009
Location: Olathe, KS

Total Topics: 12
Total Posts: 623

Posted 07/06/09 - 08:29 AM:	#7
Incision wrote: But the sets {A, ~A} {(A & B), (A & ~B), ~A} are exhaustive and constituted of members indistinguishable except for their names. Really? I may be getting in over my head here, but in your first example, it seems to me that A and ~A are distinguishable by virtue of the fact that one is the negation of the other, not just from their names. In the absence of any information whatsoever about A, it seems reasonable to assign 50% probability to both A and ~A. In your second example, the first two options are distinguishable from the third by virtue of involving (presumably) two different propositions. To satisfy PIR, you would have to replace ~A with (~A & B) and (~A & ~B). In addition, all of these variables are only indistinguishable as placeholders; once actual propositions are substituted for A and B, they become distinguishable by other means.
"Be attentive, Be intelligent, Be reasonable, Be responsible." - Bernard Lonergan (1904-1984)

aletheist
Professor

Usergroup: Sponsors
Joined: Feb 06, 2009
Location: Olathe, KS

Total Topics: 12
Total Posts: 623

Posted 07/06/09 - 08:51 AM:	#8
Incision wrote: Let P be the conjunction of all contingently true propositions. Then P itself must be a contingently true proposition. So if the PSR holds, then there should be a true proposition E which explains P. This E is either contingently true or necessarily true. But it can't be necessarily true, since if it entails P, and P is contingent, then P is necessary, which by hypothesis is false. Two questions: 1. Are explanation and entailment equivalent--i.e., if E explains P, does E necessarily entail P? 2. Could a determinist get out of this apparent bind by claiming that there are no contingently true propositions?
"Be attentive, Be intelligent, Be reasonable, Be responsible." - Bernard Lonergan (1904-1984)

Incision
Professor

Usergroup: Administrators
Joined: Jan 04, 2008
Location: Utah

Total Topics: 23
Total Posts: 906

Posted 07/06/09 - 02:16 PM:	#9
aletheist wrote: I may be getting in over my head here, but in your first example, it seems to me that A and ~A are distinguishable by virtue of the fact that one is the negation of the other, not just from their names. I suppose that's true. But notice that if they are distinguishable just because they have different relationships to each other, then no exhaustive and mutually exclusive set has members indistinguishable except for their names. For example, suppose I tried to meet your objection by getting rid of {A, ~A} and replacing it with {A, B}, ostentatiously stipulating that {A, B} is an exhaustive and mutually exclusive set -- that necessarily exactly one outcome obtains. In that case, the spirit of your objection would still have equal cogency: A and B are distinguishable because it's true that if A then ~B, but false that if B then ~B. And that's how it will work for every exhaustive and mutually exclusive set: if heads then not tails, but not if tails then not tails, so heads and tails are distinguishable. So if that's what we mean by "distinguishable," then the PIR is only vacuously true. Sure, "all" exhaustive and mutually exclusive sets, with members indistinguishable except for their names, have equally-probable members, but only because there are no such sets. I guess I was thinking that by "indistinguishable" we meant "indistinguishable in relevant ways," that is, "indistinguishable in any way that would effect probability." There's no reason why ~A should generally have a lower probability than A, so the members of {A, ~A} are indistinguishable in that sense. aletheist wrote: 1. Are explanation and entailment equivalent--i.e., if E explains P, does E necessarily entail P? 2. Could a determinist get out of this apparent bind by claiming that there are no contingently true propositions? (1) Almost. E explains P to the degree that E makes P probable (as a minimum condition -- explanations also must be noncircular [and, maybe, true]). If E materially implies P (if it's simply true that if E then P) then E gives P a probability of 1, so you could view material implication as a sort of an upper limit to explanation. (Not sure if that's clear.) And if necessarily E would give P a probability of 1, well, then E makes it completely unmysterious that P. But E doesn't have to necessarily entail P to have some explanatory power. (2) No. Suppose I said that all propositions are necessarily true -- that I couldn't have had one more hair than I have, that McCain couldn't have won the US election. Now consider the conjunction of all these necessarily true propositions. Does it have an explanation? Well, since all propositions are necessary, any explanation would have to be necessary. But then the explanation would be part of the explicandum; the explanation would be obviously circular. That means it wouldn't be a explanation at all, so by reductio we conclude that there is no explanation. Well, thanks for putting up with my long posts. I realize this is not exactly what you had planned to discuss and will try to make my comments more helpful in any future responses.

aletheist
Professor

Usergroup: Sponsors
Joined: Feb 06, 2009
Location: Olathe, KS

Total Topics: 12
Total Posts: 623

Posted 07/06/09 - 02:24 PM:	#10
Incision wrote: Well, thanks for putting up with my long posts. I realize this is not exactly what you had planned to discuss and will try to make my comments more helpful in any future responses. Not a problem! These are interesting points to ponder. Besides, it is not like I have gotten a ton of feedback on the "philosophy of engineering" aspects of the OP.
"Be attentive, Be intelligent, Be reasonable, Be responsible." - Bernard Lonergan (1904-1984)

Page: 1 2

Sorry, you don't have permission to post. Log in, or register if you haven't yet.