There has been a lot of discussion recently about different views of science, and mutual incomprehension between a certain group of students of the ways of science (varying from "postmodernist" literary critics to sociologists who have studied the ways that scientists work) and many scientists themselves, who find some or all of these views incomprehensible.
Two facets of the debate (it has many) could be put crudely thus. One facet holds that science is an activity in which people engage in a search for the truth about various aspects of a physical world which contains us; successful science is that in which the truth can be determined to all intents and purposes. There are many things left undecided (where is the dark matter?), and many that do not admit of determination (who is the Dark Lady?), so this inquiry cannot ever be expected to be complete. The other facet holds that science is essentially a political activity involving consensual agreement among members of a guild-like community (that of "scientists"): successful science is that whose proposals succeed in holding sway, and "the natural world has a small or non-existent role in the construction of scientific knowledge" (attributed to Harry Collins by Susan Haack (in Manifesto of a Passionate Moderate, University of Chicago Press, 1998, p91), also by Jean Bricmont and Alan Sokal (Science and Sociology of Science: Beyond War and Peace, Jay. A. Labinger and Harry Collins, eds, The One Culture: A Conversation About Science, University of Chicago Press, 2001). Bricmont and Sokal suggest Collins is paraphrasing a view he may not himself hold. Collins has confirmed this in a private communication. His view at the time was that SSK practioners "must act as though truth, etc ..... However, his comment was repeated in the introduction to the journal's issue without the conditional. Collins notes that his view of the relationship of evidence and truth changed thereafter changed, around 1981).
Those proposing the second facet often claim that the notion of truth is essentially relative. There are no truths as such; there are just propositions that people assent to (a view associated with Richard Rorty). This view conflicts, for example, with a claim that if we all died out tomorrow from some mysterious disease, the earth and sun would nevertheless continue in their relative motions. For if there were no humans, there could be no consensual agreement and thus no truths. Since such a view of truth conflicts with what appear to be truisms, it is taken by many to be a reductio ad absurdum of the second facet. Susan Haack suggests that "Rorty has stripped [the notions of] "justification", as he has "belief", "inquiry" and even "discussion", of essential content." (p20, Manifesto of a Passionate Moderate, University of Chicago Press, 1998).
There are cultural traits which induce the two facets into mutual incomprehension. One concerns style of writing. There is a form ideally used in science and analytic philosophy, in which, crudely put, every sentence is supposed to be literally true, and in which a thesis is declared, which is then (supposed to be) supported by rational argument (a précis of a deductive argument, let us say). Let us call this the LT form. A piece of LT writing is judged to be acceptable or not according to whether every sentence is true and the thesis contained therein is indeed rationally supported by the argument presented. At least, that's the way the story goes. Other sorts of writing, say poetry and fiction, are not LT. They adhere to a different convention in which all sentences have the same status (they are not necessarily to be judged according to their truth) and what matters is how they are linked together and how they affect an observer. These two forms of communication do not invariably handle distinct topics. Poetry is often said to convey truths about the world, and there is a real sense in which the best of it does. Conversely, one often speaks about the beauty of particular pieces of science and mathematics and compares it with music or painting.
But while few scientists might feel that Einstein's short paper on Brownian motion would qualify as an essay for The Paris Review, some adherents of the second facet may well propose that it might. They have suggested that poetry, political tracts, mathematics, experimental science, indeed all writing, have equal status. Indeed, whatever status writings may have under which they are equal, such a view does seem to follow from Rorty's claim that truth is just another word for agreement. Either way, assessing the literal truth of assertions does not appear in any case to be a major criterion for much successful writing, even discursive writing. Let us call this form in which "everything is equal" the EE form.
An EE position is associated with postmodernist deconstructivism. Deconstructionism is a name for a style of criticism in which one expresses one's reaction to a text and this reaction itself is taken as being a fact (insofar as there is any) about the text itself. Asking what an author "really meant" is regarded as a faux question. Indeed, there is no essential distinction between literature and criticism. In this milieu, science and literature are to be judged according to similar criteria, as are literary criticism and scientific commentary, themselves being literature and science, and reading novels and learning how to deal with differerential equations are to be treated equally.
It is popularly supposed, with some justification, that deconstructivists cannot give a coherent story as to how, in virtue of their writings alone, they get read and listened to in preference to the commentary of my four-year old son. But they can reply that neither can scientists give a coherent story as to how science is the way it is. They say, with some justification, that what scientists say about what they do is not actually what they do; that what they actually do has consequences not only for the development of their science, but also for the rest of society; and that these consequences cannot adequately be explained by the "rational reconstruction" of their activities that scientists would prefer to have accepted as their public face.
Not only the postmodernists can reply thus. Those sociologists who study SSK are also interested in how science is actually done, and in the discrepancies between what scientists claim they do and how their field "progresses", and what scientists actually do and how the historical record and current observation says things got and get decided or resolved. For without a doubt there are discrepancies, and one cannot say whether they matter or not unless one looks. Like Pandora, they look.
One might imagine that EE entails non-discrimination of texts. But this cannot be so. Postmodernism and deconstructivism have their journals; articles are selected on some non-random basis; tenure decisions are still made partly on the basis of publishing record (try publishing nothing!), and so on. An interesting - sociological - question concerns the publishing habits and styles in different disciplines, for they vary widely. Even though one may not believe in "standards", one can still inquire what the standards are in a particular discipline. So how about comparing standards of scientific journals with those in postmodern studies?
Enter Alan Sokal, a New York physicist who likes a prank, and who wanted to discover whether there was a demonstrable difference in intellectual judgement involved between, say, getting a paper accepted for a major physical journal and getting it accepted in a major social commentary / "cultural studies" journal. He seems to have found an example of one, although he notes he did not exactly perform a controlled experiment. He wrote a spoof article which took a number of outrageous statements about physics and society (some of them lifted verbatim from some of the deconstructionist thinkers), edited it carefully to make sure it was utter nonsense, and sent it to the renowned journal Social Text. It was accepted and published in 1996 without peer review, on the judgement of the two editors who admitted in print later that they didn't understand it. Sokal published an article a little later in the journal Lingua Franca in which he announced it was a hoax, and explained why he had done it, namely to show that intellectual standards in the target milieau were vapid. His physicist colleague Jean Bricmont joined his side of the debate.
Then the outraged hordes descended (pick your side). Sokal and Bricmont published a book, (op.cit.) which attempts to deconstruct deconstructivist writing and leaves the remaining characters in a little tangled heap on the floor. A number of prominent intellectuals, mostly French, were taken apart chapter by chapter. Sokal and Bricmont appear to have been living high on the debating circuit ever since. As well as their book, they have published a large number of papers on the topic in the haunts of the litcrit and SSK crowds. They do seem to have gone native. Well, one might think, somebody had to.
My impression is that this topic is far more complicated than it seems. The views of the radical relativists are hard to refute on their own terms. EE views may be self-consistent not only with respect to EE standards but also with respect to LT form. Self-consistency is, however, not the only criterion of acceptability (although it has been proposed to be so in mathematics, for example, in the Hilbert program). Bricmont and Sokal claim that this self-consistency is bought at an epistomological price far too high reasonably to pay.
One suspects that the editors of Social Text felt that Sokal's disdain had a smidgin of truth (sorry, consensus) to it. They could have argued, consistently with EE and deconstructivism, that
Bricmont and Sokal have two main theses. The first is my main topic, discussed later. The second is that they consider much of "postmodernist" deconstructionist-style writing intellectually vapid and establish this claim, as do others, by quoting at length. They do find some wonderful and amusing tidbits, as do others such as Susan Haack (op.cit.).
If one is judging according to the LT criterion, it is appropriate to critique quotations which appear to be meaningless, false or fanciful; every sentence is supposed to be true, and any that is not is ground for criticism. One may think to contrast LT with, say, literary criticism, in which quoting a few lines of Thomas Mann would barely suffice for a serious commentary on any novel of his. However, it is difficult to formulate LT precisely. Boundaries become blurred. Even in LT form, some sentences must be considered in context (for example, sentences asserted as consequences of previously-introduced counterfactual hypotheses could not be taken to be literally true). An adherent of EE could point out that even LT adherents acknowledge the importance of context in evaluating (what they claim to be) truth, and so a series of arguments supported by selected quotes is just a further deconstructivist contribution which by no means invalidates the original text, as Sokal and Bricmont hope to do.
One notes further that EE provides a uniform theory for all forms of written communication. In contrast, the status of literary writing has long posed a philosophical puzzle for those whose realist philosophical predilections incline them to LT. While having a theory of everything is not always preferable to having a theory of some small thing, even some scientists claim it as a selection criterion amongst their own theories.
Bricmont and Sokal's attempt to deconstruct the deconstructivists, Impostures Intellectuelles, takes people to task. Julia Kristeva, known inter alia for her biography of Hannah Arendt, and one of Bricmont and Sokal's targets, is reported by the newsletter of my university to have declined an invitation to come to the ZiF research center at Bielefeld to debate Bricmont and Sokal and others, on the grounds that she considered it - whether the invitation or the debate is not clear - an insult to "French culture". How may we take this?
One response might be to note that the situation as described is as cute a comment about what "French culture" may take itself to consist in as any by Bricmont and Sokal. It seems that not only Groucho Marx considers an invitation to speak one's mind as an insult to one's culture. A second response might be to suppose that the situation may have been misdescribed by the newspaper, due to misunderstanding or even bias. There could be some simple animosity between the participants that would preclude an interesting debate, for example, and Madame Kristeva could have known so.
What can we make of these two responses? Surprisingly, the first is a passable deconstructivist response to the pure "text". I read it: it exists; I am entitled to comment as it occurs to me to do so, and what occurs to me is Groucho's disdain for being a member of any club willing to offer him membership. Basta. I am part of the "development" of the text and Madame Kristeva, the Bielefelder Universit\"atszeitung, Peter Ladkin and Groucho Marx are forever linked in literary history. What a rich world!
The second response supposes there was a fact of the matter about what went on, and the report partially but accurately reflected that fact of the matter, or it did not. That "fact of the matter" involves a lot of uncertainty, however. Imputations of emotions, responses, the meaning of certain types of exchanges, intentions, and so on, many of which are open to doubt and likely remain undiscoverable after the fact. Social and psychological "facts" are notoriously slippery things on which to get purchase. For all our presumptions, we may never be able to decide between the proposal that Madame Kristeva's reply was generated by her thoughtful self, and the proposal that she has memorised a few thousand sentences, selects one randomly as the basis for her reply, and writes supporting material around it. We simply have no way of really knowing. All we have that we can really rely on is what she wrote, and what the university newspaper wrote. Any further interpretation is, in a real sense, ours. Apart from the existence of the text, any "fact" comes ultimately out of our own heads. So the "deconstructivist" comment involving Groucho has, in some sense, more reality to it than any supposition I may make about what "really went on" during an exchange of invitations, even assuming the newspaper wasn't making the whole thing up as some sort of late-summer April Fool's.
So is this all, as a prominent US intellectual is reported to have said, "French nonsense! Gibberish!"? Let us put it in the philosophical context. Western modern philosophy has tried for centuries to construct the world out of signals impinging on the senses, out of so-called sense data (from the British empiricists to Husserl). In response to scepticism that this could effectively be done (Hume's comments on the epistomology of causality), intellectual frameworks were constructed in which our modes of comprehension of the world are inevitable (Kant). Then, after many further adventures, it was proposed to explain the world on the basis of the ontology of logic (the logical atomists: Russell, Wittgenstein) and veridical observations of the world were supposed to be limited to these (the logical positivists). Then positivism was rubbished. And so on. Through all of this, the notion of how descriptions relate to the things they purport to be describing, of how assertions represented in language relate to the "facts" in the world they are supposed to describe, remains highly problematic.
Deconstructivist postmoderns start from text, from what has been said, which is at least somewhat more concrete than sense-data ever were (it is hard coherently to doubt the existence of texts, unless one is a radical Cartesian sceptic; in contrast, even today people wonder what sense-data might be), and although they may construct an infrastructure on top of that, they do not necessarily attempt to construct reality and "facts" the way that Western empiricists or rationalists have attempted. For example, they have the same kinds of difficulties reconstructing the notion of truth that even true believers in the idea have. But rather than institute a social convention to believe in LT and figure out the details later, they have instituted a social convention according to which, if one cannot figure out what truth consists in, then one does not presume to judge complex writing according to it.
That has a ring of practical sense to it. It is hardly a radical sceptical position, which, as Bricmont and Sokal point out, seems to have little practical sense to it at all. It seems to be consistent with all sorts of things, including the Quine-Duhemian and Kuhnian orthodoxies in foundations of natural science. And it takes twentieth-century analytic philosophy seriously in putting language at the center of things.
It is popularly supposed that believing in the relativity of truth and thus of facts leads one to have difficulty crossing the street (one is taken to be a sceptic about the existence of cars and social rules of the road). The persistence of the form shows this worry to be misplaced, for believers do not appear to be selected against naturally, as Darwin might have supposed. Hume didn't have trouble crossing the street either, despite his purported scepticism about causality (although some recent commentators have taken Hume's to be an epistomological position concerning ignorance, and not a sceptical position vis a vis causality). Crossing the road, like dunking witches, is probably not a good criterion for intellectual veridicality. Many philosophers have considered starting from language as the basis for reconstructing whatever it is that our reality consists in, indeed some consider this the basis for twentieth century analytic philosophy. One can hardly consistently begrudge the deconstructivists their fundament. But what about the direction in which it takes them? Maybe EE is another Wittgensteinian language-game whose rules are obscure to most Anglo-Saxon analytic philosophers, most professional scientists (though not all), and most mathematicians. Wittgenstein persuaded us that there are no rules about what rules there are, not even this one. The particular game played by deconstructivists seems to be more kin to performance art than it does to the stuff that I like to look and wonder at in art exhibitions.
That all said, does a working scientist have to engage deconstructivism in daily life? No. A working philosopher of science? No. Maybe a computer user or a bicycle rider? No. Taxi driver, bus driver, pilot? No. It seems as if many of us can get along fine without worrying, as long as the litcriterati don't outpatois my wine dealer and run off with all the good wine. Or the research grants which belong to Truly Serious People such as, well, myself.
So the relativist enterprise itself does not appear to be out to lunch, though some of the litcritical targets of Bricmont and Sokal's commentary may appear to be. But therein lurks a Trojan Horse. Bruno Latour is a sociologist who has made a career from sneaking into labs in broad daylight and publishing details about what consenting scientists do to each other, not all of which may be suitable for children.
Bricmont and Sokal find enough gibberitical statements of sufficient brevity in Professor Latour's work to haul him over the coals. However, Latour is also renowned, even revered, and certainly rewarded, as a perceptive and original commentator on the ethnography of scientific activity. Scientists, and philosophers of science, must take his work very seriously indeed. Among other things, he is one who can tell the governments (of three countries at least) what we actually do with its money rather than what we say we do.
That is not a totally facetious comment. The philosopher James Fetzer took a prominent swipe at the formal verification community of computer scientists in 1988, in what was then one of the major organs of computing science and technology, the Communications of the Association for Computing Machinery, in which he claimed to have demonstrated that program verification was impossible, and suggested that funding for it should be cut. This engendered howls of outrage from amongst the formal verification community, whose major published arguments in response appealed to authority to judge Fetzer's argument to be nonsense. Unfortunately, their appeal to authority was self-reflexive and thus not exactly persuasive to those outside the community (or even some of those within it, such myself and Leslie Lamport, who noted to the authors that "Fetzer made mincemeat of you, and Denning ground the remains into the dust", quoted in Chapter 6 of Donald McKenzie, Mechanizing Proof, MIT Press, 2001, which recounts the controversy). Nobody seemed to address the substance of Fetzer's arguments, although two later comments (by Brian Randell and John Dobson, respectively, Jon Barwise) went some way towards it. Oddly, Fetzer's arguments yield to straightforward counterexamples, and it remains a mystery to me why these were not regarded as decisive refutations of his view. Some computer scientists have also said to me in private that they thought he was largely correct. It seems there is no uniform view in the computer science community about the correctness of Fetzer's arguments. At least they can't be considered to be knowledge, consensual or otherwise.
So I was already disposed to be curious about the underlying sociology before the discussion over Sokal's spoof. It seems as though there are issus to get serious about. Let us engage.
The "Sociology of Scientific Knowledge" debate is intellectually non-trivial. "Knowledge" in SSK means not what it does to philosophical epistomologists. Philosophical epistomologists used to consider knowledge as something like "justified true belief" (no longer, since the Gettier examples. Maybe knowledge is something like a true belief that is justified by that which explains its truth, but I don't mean to pursue this here). "Knowledge" in SSK means "shared institutionalized belief" (cf. Donald MacKenzie, Knowing Machines, MIT Press 1996/1998, p9).
Barry Barnes and David Bloor delineated the "Strong Program" (SP) of SSK (Barry Barnes and David Bloor, Relativism, rationalism and the sociology of knowledge, in Martin Hollis and Steven Lukes (ed.), Rationality and Relativism, Oxford: Basil Blackwell, 1981), in which the practioner is enjoined to pursue hisher researches without being influenced by the actual or claimed truth or falsity of the scientific claims which are the subject of the subjects of the study. SP is strictly agnostic about what the truth is; it is a negative thesis concerning the relevance of truth. Holding a relativist theory of truth, and being agnostic about truth are not the same thing, but a tendency to conflate them may have led to some misunderstanding. Especially since some sociologists do seem to accept a apparent relativist idea of facts (thus the comment cited above: "the natural world has a small or non-existent role in the construction of scientific knowledge").
Many scientists do not accept, though, that agnosticism about truth is an appropriate position to take for those describing their activities. These scientists believe that one cannot tell a coherent story about scientific activity unless one incorporates information about the correctness - meaning truth - or otherwise of the scientific views. This is what is thought to give science its prestige. The scientific way (whatever it might be) is regarded by its practioners as a demonstrated efficient set of ways to ascertain the truth about features of the physical world, and if you take the "truth" bit out, there is concern that you could't distinguish it from any religion. Indeed, the philosopher of science Paul Feyerabend argued some thirty years ago that there was nothing essential about science to distinguish it from other myths. Some regard Feyerabend as having taken the Quine-Duhemian and Kuhnian orthodoxies to their logical conclusion. But this kind of conclusion won't do, say some scientists. Galileo and the pope's people discussed substantial issues: does Jupiter have moons or doesn't it? This kind of question is not directly comparable with attempting to determine the identity of Shakespeare's Dark Lady. The former is, well, factual, and the latter not. Darwin and creationists had similar factual disagreements (and it is still going on 150 years later). It isn't just a football game. The earth isn't flat, and nobody can coherently say that it is. Jupiter really does have moons, as Galileo said (note that it took a few hundred years for the church to reconcile itself with Galileo, whereas it takes much less time for scientists to accept new discoveries that contradict their own work). No religion has built airplanes, immunized us from smallpox, measles and poliomylytis, extended our life expectancy to 75+ years, or landed us on the moon. No other "myth" has come anywhere near such distinctive consequences.
A weaker form of objection to thorough-going SP would consist in suggesting that sometimes, not necessarily always, the truth or falsity of some scientific proposition or other is relevant to a sociological account of the science. I am amongst those who incline to this view. Indeed, some sociologists may incline to this view also.
However, there are good reasons for not factoring in truth, sociologists can reply, so whether or not thorough-going SP is achievable, it should at least be attempted. The exhortation to SP contains no proof that SP is always achievable: it is a method, not a fact.
Justification for this method is as follows. First, that something is true is almost always no explanatory help at all in figuring out how people came to accept its truth. Knowing that Fermat's Last Theorem is true, as most mathematicians believed for a long time, is no help at all in describing Wiles's and others' struggles to establish its truth, and that latter is what we sociologists are interested in. And if a flaw is discovered in the proof in a couple of years or decades, should that invalidate our social observations on Wiles's discovery? If doesn't seem as if it should, but if we relied essentially on veridicality or objective correctness to make our observations, it would.
Consider another example: Hilbert's formalist program to demonstrate the rationality of mathematics by exhibiting methods of reducing mathematical statements to those of simple arithmetic. Such a reduction, if it succeeded, would then show that specific mathematical theories were self-consistent if arithmetic were consistent, and Hilbert suggested that self-consistency was sufficient for the validity of mathematics (we have already seen that self-consistency is not sufficient for, say the validity of LT or EE, so this would have been a singular and contentful view of mathematics, had it succeeded). Hilbert's program foundered on the discovery by Gödel that arithmetic could not be shown to be consistent, except by methods that were logically stronger than arithmetic itself, unless arithmetic were itself inconsistent. This meant that if one pursued Hilbert's reductivist program, it would have to be based on an assumption that arithmetic were consistent, for there could be no elementary proof that did not involve a logically stronger theory. But further, no theory that logically implied the consistency of arithmetic could be reduced to arithmetic, as Hilbert wanted, unless arithmetic were inconsistent and thus a useless basis on which to establish the rationality of mathematics. Such is the extent of the mathematical facts of the matter. It is of no help in explaining how this knowledge progressed to judge the actions of the participants in Hilbert's program to have been pursuing chimeras. They were not eccentric. They just lacked knowledge and proceeded on the basis of a mistaken but explicit assumption. In explaining what they were doing, it may be helpful to suspend one's hindsight into the mathematical facts as they turned out to be.
These are examples from mathematics, but the point applies more widely. If scientists are attempting to work out the answer to something, then they don't actually know it at the time themselves, so in describing what they are doing, appealing to truth or falsity of propositions would be hindsight functioning as a deus ex machina and lead us away from accurate descriptions of the social facts. (This general point has been attributed to Harry Collins.) This argues for SP.
One can imagine a contrary response thus (call it CR). If you are looking at a process in which everybody eventually comes to the conclusion that 2+2=5, you surely have a lot of different explaining to do than if you are describing a similar situation in which the received knowledge becomes 2+2=4. In the first case, that everyone was seriously misled is a social fact that needs to be explained, whereas a comparable social fact is absent in the second case. The cases are not even sociologically analogous. But if you ignore the truth or falsity of what is believed, then you cannot differentiate the two cases.
In reply to CR, a truth-relativist might query whether there is a social fact there at all that needs to be explained. Heshe might argue that the claim that there is a social fact to explain is dependent upon a judgement of the truth or falsity of the beliefs. If one believes that that truth or falsity is conventional, then one cannot establish that there is a fact here to explain. Thus CR depends upon an assumption which a truth-relativist can deny. Is there maybe another way to establish an argument using an unequivocal sociological fact which differs according to the truth or falsity of a statement? Let me try to find one.
There are two steps. The second step implies the existence of partly commensurable, partly incompatible scientific theories. Some may think that there cannot be such things, maybe invoking Kuhn's work on the structure of scientific revolutions, that when a truly new theory of sufficient explanatory strength appears, the new theory and the old theory are strictly incomparable. So the first step is to establish that such a situation of partial comparability of incompatible theories can indeed pertain, or, at least, that there is a good argument that it can. I contend that there are many ways to compare logically incompatible theories (that is, two theories, each of which has a consequence whose negation is a consequence of the other). Recent work by Friedman (Dynamics of Reason, CSLI Publications, Center for the Study of Language and Information, Stanford University, 2001) has highlighted comparabilities between the "classical" examples of incomparability, Newtonian and Einsteinian theories of mechanics and gravitation (Friedman refers to Cartan's technical comparability work in 1923-4, in which he established Newtonian mechanics within a mathematical framework of the same sort to that of relativity theory). Work on logics has shown many degrees of comparability of the prima facie incompatible principles of deductive reasoning espoused by intuitionist mathematicians and "classical" mathematicians. I shall leave it there, but some more detail concerning partial comparability may be found in the appendix.
For the second step, I now attempt to construct two partially incommensurable theories and differentiate them through a sociological fact which is functionally dependent on the truth or falsity of a given logical assertion.
The "Dutch Book Theorem" (Michael Resnick, Choices: An Introduction to Decision Theory, University of Minnesota Press, 1987) says that if people mistakenly assess likelihoods, then one can make a "Dutch book" against them and run off with all their money. A Dutch book is a bet which the offerer cannot lose, no matter which way affairs turn out. The Dutch Book Theorem is taken to be a way of showing that those who interpret likelihoods as determined by which bets one considers equivalent (an interpretation often attributed to Frank Ramsey) are committed to the Kolmogorov axioms of probability. So it somehow connects likelihood judgements with social consequences.
I suppose a situation in which communities A and B agree on certain elementary facts concerning arithmetic, but community A believes certain types of physical falsehood, and community B does not. (I note that my supposition implies that partial incompatibility of theories must be possible, hence my need to establish that as the first point.) The truths on which communities A and B differ must also not be epistomologically certain. There must be some residual doubt. So community A believes a falsehood X (with some residual doubt) and community B believes instead the truth not-X (with comparable residual doubt).
I suppose agreement between A and B on enough arithmetic that they can agree on calculations regarding the settling of bets. I also suppose agreement between communities A and B on the desciption of the outcomes of certain experiments, namely those which can be devised to test X, resp. not-X.
If these suppositions are fulfilled, community B can now make a Dutch Book against A. A believes it can make a Dutch Book against B, but in fact community A will collectively lose its shirts because not-X is true. And losing your shirt is a social fact. The connection I was aiming for is established.
Or at least I hope so. There are some details to be filled out in this argument. For example, it only works for some kinds of physical truths with simple experimental manifestation; say, laws implying some value for the frequency of an observable phenomenon. And I haven't given an example of one. But it does give one way of attempting to show how physical truths may sometimes be differentially convertible into social facts. How does this bear on SSK's principle of ignoring truth (but not truth claims)?
A naive realist could claim contra SP that the argument shows that truth or falsity of beliefs, namely the ones fulfilling the assumptions above, necessarily impinges on social issues in a way that even truth-agnostic sociologists may not ignore, and thus that SP is inappropriate. An SSK practioner of SP could reply to the contrary that since the distinguishing feature is social, this puts the claim X (resp. not-X) well within the bounds of social commentary, and that as a consequence of believing X, community A lost their shirts. This would show, not that SP was inappropriate, but only at most that it is redundant, for if many physical truths can be converted into social facts, then paying attention to the social facts must entail paying attention to these truths, which was the very point that the naive realists were trying to establish. So there is no argument contra SP here.
A truth-relavist could go even further, and argue that the salient fact is not that X is a physical truth, but that communities A and B agreed on the arithmetic used in calculations of the bets. And who knows whether this arithmetic is necessarily the same as our arithmetic? Maybe A and B could agree, through use of a different arithmetic, that B loses its shirts instead. If this is so, then it would seem that A lost its shirts through agreement on a certain type of arithmetic, rather than through the fact of X being or not being true. So agreement is once again paramount.
I hope at least to have illustrated that the issue of the applicability of SP is not trivial. Even with a prima facie strong connection between physical truths and social facts as illustrated by the argument through a Dutch Book, it remains possible to hold that some of the salient considerations are nevertheless social, and that what matters is agreement and not the logical truth or falsity of the arithmetic that led to that agreement.
The final argument for SP which I wish to note is that SSK practitioners claim that they are not expert scientists in the fields they study, and are thus not qualified to say what counts as the truth or not; further, even scientists change their minds after a couple of years or a couple of hundred years (this point may also be attributed to Harry Collins). To which I guess the scientist can reply: if you are disqualifying yourself on grounds of intellectual incapability, then if you had that capability would you use it? If not (and the Strong Program suggests you should not), then lack of qualification cannot be the main reason for truth-agnosticism. If it would be used if available, then one is simply pleading ignorance, and SP is just a way of making the best of a bad situation, and not by itself a justifiable principle of SSK methodology. I do not consider the argument to SP from lack of qualification to be very strong, despite its intuitive plausibility.
The bottom line is probably something like this. SP or not, each piece of SSK work stands and falls on its own, as to how well it exhibits and explains the social facts in a particular scientific endeavor. Thus Steven Shapin (How to Be Antiscientific, in Labinger and Collins, op. cit.), if I may paraphrase loosely: "I do what I do, and anybody who wants to criticise my work is welcome to get into the details and show me how what I have said is incorrect, if it is." This is an appropriate challenge to any critic. It says that a critique of SP must apply to individual cases if it is to be pertinent. Surely this is correct. It remains only to note that it contains no general argument for or against SP.
The activity of natural scientists has observers besides those scientists themselves. Those observers see many different things. Science journalists and scientists themselves see discoveries and advances in knowledge. Sociologists of scientific knowledge see less of the natural world and more of social constructions of claims, and so-called postmodern critics in cultural theory seem to see virtually nothing of the natural world, and see nothing in science to distinguish it from, say politics or literary criticism. I am, or would like to be, some variety of naive realist about science, as are most scientists, but I do lean towards a view captured well by the parable of the blind men with the elephant. The blind men feeling the trunk, the legs, the tail, and the ears all perceive very different things. Nevertheless, they are all parts of the elephant. And it remains true that the elephant has one trunk, not two, two ears, not four, and four legs, not eight.
Many physical scientists view science as epistemically privileged. The activities of science leads towards knowledge of the world that is, according to this view, much more well founded than that of, say, Marx about the prime causal factors affecting society. In contrast, as we have seen, some postmodern cultural commentators have denied it, and others, from sociologists of scientific knowledge to philosophers of science and even some scientists, have doubted it. It may be well at this point to ask where the views of various scientists originate.
The philosopher Larry Laudan calls the attempt to show science as epistemically privileged the demarcation problem, and has encapsulated the historical development of the demarcation problem thus (Laudan, The Demise of the Demarcation Problem, Chapter 11 of Beyond Positivism and Relativism: theory, method and evidence, University of Chicago Press, 1996). Aristotle said that knowledge may be called scientific if it is apodictically certain, and that the way to certainty led through identifying causes, using deductive reasoning, and identifying universals which "inhere" in the particulars of sense. He also distinguished between "knowing how" (the shipbuilder knows the characteristics of ships, and how to build one with those characteristics) and "knowing why" (a scientist knows why the ship floats, and why it behaves as it does when built to a particular form by the shipbuilder). By the time of Ptolemy, many had given up the attempt to derive the motions of the planets causally, and the Ptolemaic system focused on describing the phenomena of astronomical motion accurately. Galileo, Huygens and Newton all advocated theories whose primary aim was to describe the phenomena, rather than identify Aristotelian primary causes (apparently Newton wished he could have given a causal explanation, but believed it was beyond his means). Nevertheless, science was regarded still as apodictically certain (a view also to be found in Bacon, Locke, Leibniz, Descartes, and Kant, according to Laudan).
In the nineteenth century, a fallibilist view of science emerged, in which scientific theories are "corrigible and may be subject to serious emendation". It became no longer possible to distinguish science as certain knowledge, and attempts were made to demarcate science through its methodology. Thinkers such as Comte, Bain, Jevons, Helmholtz and Mach tried to identify something called the "scientific method", which, although fallible, was somehow self-correcting and thus a reliable way to knowledge of the world. There were various views as to what this privileged method was. Herschel and Mill took it to be the use of so-called inductive reasoning. Others took it to be the restriction of theories to observable phenomena. Others (Laudan cites Whewell and Peirce) to it to be the ability of theories to make surprising (true) predictions. Also, many of the proposed specific methodological rules were ambiguous; it was hard to tell when they were being followed and when breached. Duhem pointed out in 1908 that many of the rules proposed to encapsulate the "scientific method" had little to do with the everyday working practices of scientists.
In the twentieth century, both verificationist (from logical positivism) and falsificationist (Popperian) demarcation criteria wer proposed, but both have been found wanting, not least because a lot of what most scientists regard as pseudoscience satisfy these criteria (Laudan notes memorably that "flat eathers, biblical creationists, proponents of laetrile or orgone boxes, Uri Geller devotees, Bermuda Triangulators, circle squarers, Lysenkoists, charioteers of the gods, perpetuum mobile builders, Big Foot searchers, Loch Nessians, faith healers, polywater dabblers, Rosicrucians, the-world-is-about-to-enders, primal screamers, water diviners, magicians, and astronomers all turn out to be scientific on Popper's criterion...."). The twentieth century brought many other views of science, such as the Quine-Duhem thesis, which essentially denied the existence of demarcation criteria. Since my purpose here is to elaborate demarcation criteria, I do not consider these views, which may even have been in the majority.
Laudan suggests some other possible candidates for demarcation criteria, to wit:
It might appear to some scientists from this conclusion that Laudan is somehow "against science", but this is not so. Laudan has criticised the "received" twentieth century views of science proposed by Quine (a holistic view, in which any individual "theoretical" claim may by maintained, if so wished, in the face of prima facie contrary evidence by giving up, or adjusting, other parts of one's entire scheme of beliefs) and Kuhn (in which major scientific theory changes, say between Newtonian mechanics and relativity theory, result in theories which stictly cannot be compared with each other in terms of what they say about how the world is; such a change cannot be claimed to be an "advance" therefore). He has also criticised the "scientific" approach proposed for the sociology of scientific knowledge by David Bloor (in Knowledge and Social Imagery, London: Routledge and Kegan Paul, 1976; 2nd edition University of Chicago Press, 1991), not least because he proposes certain criteria for being scientific; in effect, he proposes a particular solution to the demarcation problem, through four criteria, three of which Laudan finds unexceptionable, even platitudinous, and the fourth of which he confutes.
I take Laudan as suggesting that scientific work stands and falls on the quality of its justification, and that the criteria for justification, as far as anyone can tell, apply to all varieties of knowledge, and not merely to those traditionally singled out as being scientific. The epistomologist Susan Haack, in writing on the debates about science and cultural theory, has also emphasised criteria of evidential quality as constitutive of true inquiry, and is less concerned with whether such inquiry is labelled as "scientific" or not (various essays in Memoirs of a Passionate Moderate, University of Chicago Press, 1998). Similarly, the philosopher of science and mathematics, Philip Kitcher, who has worked on potential demarcation criteria for science (see his critique of creationism, Abusing Science, MIT Press, 1982), has focused recently on the actual arguments used by scientists (notably Darwin and Lavoisier) and why they were successful (in The Advancement of Science: Science without Legend, Objectivity without Illusions, Oxford University Press, 1993). There seems to be a consensus here, despite individual differences, that what counts is whether certain beliefs, certain propositions about the world, are well justified or not, and not whether something is labelled as "scientific" or not. Laudan points out (Science at the Bar -- Causes for Concern, Chapter 12 of op.cit.) that biblical creationism is refuted, not by being shown to be "unscientific", but by being shown to lack justification; it is refutable, as Popper might wish, and also refuted. One might suggest that creationism is simply bad science. (This has the interesting implication, not that it should not be considered in school curricula, but that it could fruitfully be considered - and be shown to be wrong. I find that such a view has a lot to recommend it. It would be a way of showing students how to distinguish good science from bad science at an early age, and maybe be an improvement over current natural scientific curricula, which, in my experience, when they deal with it at all, seem to hold to more traditional demarcation criteria, which I think Laudan and others have shown to be lacking).
Many physical scientists seem to adhere to a view of science as epistemically privileged, and that there exist demarcation criteria. At the same time, most scientists recognise that there is good science and bad science, and there is substantial consensus as to which is which. I think it may be fair to say that good science is indeed epistemically privileged because well justified - indeed, those might be two ways of saying the same thing. However, what goes for science also goes for views on science as an activity. Many scientists applaud the efforts of Jean Bricmont and Alan Sokal to defend science against what they see as misguided criticism (from sociologists as well as postmodern cultural critics). They have inveighed against what they see as poor intellectual standards amongst critics of science, as compared with those of scientists. They argue, in the words of this section, that many of those views, although famous, are not epistemically well justified (actually, that they are nonsense). But they have also engaged with some substantial theses about scientific activity proposed by sociologists of scientific knowledge and have claimed they are false. It appears superficially plausible to many scientists to consider these theses as false, but I believe they are largely accurate characterisations of the everyday business of doing science, and now is the time to illustrate why I think so. My illustration falls far short of a proof, however. It consists largely of a set of anecdotes and examples from my experience, along with a presumption that my experience is not so very different from that of others.
I think Bricmont and Sokal's critique of the Strong Program (SP) is weak, because I think the views of science that they criticise, understood appropriately, are mostly justified when one considers everyday practice in science and the results of that practice. However, I understand those views against which Bricmont and Sokal argue maybe differently from the way they might.
Bricmont and Sokal suggest (in Science and Sociology of Science: Beyond War and Peace, in Labinger and Collins, op.cit.) that the SP position is comparable with radical Cartesian scepticism. I don't think this is so. Neither do I think it is correct to deny the specific theses which they claim SP adherents are committed to. I think one can be even a naive realist about science, far from a sceptic, and still support the theses which Bricmont and Sokal criticise in their argument that SSK is radically sceptic, with the exception of the thesis of the relativity of truth.
The specific view that I want to critique is as follows. Bricmont and Sokal argue that the theses of SSK, that all facts are "socially constructed", that scientific theories are "myths" or "narrations", that scientific debates are resolved by "rhetoric" or by "enlisting allies", (as they list in Labinger and Collins, op.cit.), are mistaken. They claim that these theses, with the addition of the thesis of truth as intersubjective agreement, are equivalent to radical scepticism. I shall argue that these views attributed to SSK are largely justified. But while I suspect Bricmont and Sokal believe they are theses which are argued to follow from, for example, truth-relativism, I would suggest that they are observations about the way we scientists and technologists go about our business.
Given that they are observations about how science is conducted, I believe that it is largely impossible to prove or to disprove such views, which does not make them any the less true. Someone who accepts a truth-relativism might be able to argue that all science is of necessity of that nature, which I take to be a much stronger view. However, I think truth relativism is false, so I am not that interested in arguments from it to other positions, for it can tell me nothing about the world. In any case, I believe that SP SSK is agnostic on truth. It has no theory of truth; a fortiori it is not committed to a relativistic theory of truth (despite occasional appearances to the contrary), so any argument from truth-relativism to these views that Bricmont and Sokal criticise need not be part of SSK's argument for them. The observation that SP is not committed to a theory of truth does not of course say anything about what theory of truth individual SSK practitioners may believe.
My commentary proceeds under the following suppositions of naive realism. There are such things as facts, and facts of the matter. 2+2=4 is a fact. There are names, and objects, and reference is real: "the sun" refers to the sun; "the moon" refers to the moon. And there are facts about them: "the sun and the earth have approximately thus-and-such relative motion" is true, or it is not the case that it is true (depending on what thus-and-such is). It is true that there is such a thing as the room I am sitting and typing in at the moment, and this room is situated on the edge of what is now a meadow. And this is true even if there is nobody around to engage in intersubjective agreement. Furthermore, there are reliable ways of inferring truths from other truths, namely the ways described by deductive logic, and what we call first-order predicate calculus is part (that does not deal with modalities) of deductive logic. And so on.
It seems appropriate to deal first with the relation of SP SSK to theories of truth. Bricmont and Sokal suggest that proponents of the SSK program adhere to a theory of truth by convention (or "intersubjective agreement", as they formulate it). As I said above, I do not believe this need be the case. As far as concerns truth theories, all that is required to perform strong-program SSK is to avoid basing conclusions on an assessment of the literal truth of scientific claims, or to use the truth or falsity of scientific claims as an explanatory feature for certain social processes. It is not required to claim any specific theory of truth. Indeed, if any piece of substantial work depended essentially on a claim that truth is convention (that is, its reasoning would be falsified, and the conclusions it draws vitiated, if truth were not to be merely conventional), then that itself would be real reason to doubt its validity as a piece of SP SSK, for it would be reasoning that depended for its validity on substantial claims about truth, and that is precisely what Strong Program SSK explicitly disavows. It seems to me rather that SP is committed to agnosticism about the nature of truth.
In any case, it is not clear to me that adherence to such a view of truth as "intersubjective agreement" must commit one to radical scepticism. For example, the theory of truth proposed by John Austin (Truth, Proc. Arist. Soc. Supp. Vol. XXIV, 1950, reprinted variously), establishes ascriptions of truth to assertions in so far as they cohere with or agree with other assertions in our complete held-true set. The notion of coherence presupposes unavoidable structural constraints (e.g., a principle of non-contradiction, I suppose), as might truth by "intersubjective agreement", if only because there is intersubjective agreement on classical logic and certain properties of truth. There seem to be many respects in which considering truth a matter for convention or for intersubjective agreement is similar to the truth-as-coherence proposal of Austin. This does not make Austin right about truth, but no one in the last half century has argued that his view commits him to a radical scepticism. Indeed, Austin's views about perception were closer to naive realism than, say, those of Locke. It seems to me unlikely that truth by convention must inevitably entail radical scepticism.
As I have said, my reading of the SP has it suggesting something like "don't use what you don't need". This is actually a position very similar to that which led to great achievements in science itself. Philip Kitcher (in The Advancement of Science, Oxford University Press, 1993) shows convincingly how Darwin's explanatory revolution gained its power. Kitcher shows that the ideas which Darwin singled out were already present and accepted in the work of his contemporaries and precursors. Darwin proposed, however, that this limited set of ideas was sufficient by itself to explain observed facts about speciation, and gave significant examples of the application of his suggested new paradigm. Creationist mechanisms, which were then common if not universal, formed no part of his limited set. Similarly, SSK can claim explanatory power if the explanations are forthcoming without referring to the truth of the propositions being investigated by the subjects of study.
Let us move on to the other claims about the social nature of scientific work. Under the naive realist view I am proposing, these claims about the nature of science are generalisations, rather like saying that the inhabitants of England communicate in English. Most of them do, better or less well, and the presence of inhabitants who cannot use English is not regarded as a counterexample to the generalisation. Rather than say that such generalisations are true or not true, it might be preferable to speak of them as appropriate or inappropriate. Thus I wish to argue that the generalisations about the social nature of science against which Bricmont and Sokal are appropriate.
To show how I mean to consider these views, let me return to the comment quoted above that "the natural world has a small or non-existent role in the construction of scientific knowledge". This may seem to be a good example of an extreme view, but let me argue to the contrary that it is appropriate. I have been dabblingin physics recently. In the last fifteen months. I attempted to construct a rigorous demonstration that a fire in Swissair 111 could not have been started alone by high-intensity radiation fields outside the airplane. I looked at the reasoning and evidence behind the claim that so-called single event upsets in digital avionics at altitude was primarily due to atmospheric neutrons, and concluded that it was very weak (which is not to say that the claim is untrue; just that no one can yet claim they know, even if they think they do). Finally, I concluded that there probably cannot be decisive epidemiological evidence for any health effects of low-dose radiation from particle decay, given the constraints upon the kinds of data one can gather. These are all very physical, practical issues. But I did all this work without leaving the confines of my office, my home office, and the university library. So I must indeed say that observations of the natural world had a very minor role to play, if any at all, in the construction of this particular knowledge. And, as postmodern critics might suspect, reading and interpreting had a very large role to play. Given what I know of the work habits of my colleagues in theoretical physics, and what they have in their offices (mostly computers, books and paper), it must be largely true of them also that they work like this. They explain or attempt to explain experimental results, but mostly they run computer simulations or do mathematics by hand.
So how big a role does looking at the natural world play, as contrasted with, say, social entities such as books, conversations, and use of artifacts? I considered the following thought experiment. Suppose I were to take the entire physics community in Bielefeld, including experimentalists, plus an oracle from God. The oracle's job is to intervene in every experiment. Every time that an experiment starts, he disconnects it from the natural world without affecting the measuring apparatus, and then he simply injects the appropriate readings into the devices at the appropriate time as they would have been injected by the natural world, had it not been disconnected. So the world can have a rest; it is being perfectly simulated by the oracle. All the physicist colleagues work as they do, and there are lots of them (a couple of hundred or so). How hard does the oracle work? I figure one or two oracles could get the job done. Now, the oracles substitute exactly for the interface between the natural world and the human social world. And it seems that there isn't much of it. Two oracles compared with two hundred physicists. One or two or a few per cent of the total work. So in terms of what physicists actually do with their time, dealing with the natural world interface ranks pretty low. However, I wouldn't go so far as to say it is "non-existent". I'd guess a few per cent.
The point of the thought experiment is this. An assertion which looks wild at first sight turns out to be a fairly good approximation to what actually happens if one looks at it in the right way. I shall now try to look "the right way", or at least a right way, at the other views that Bricmont and Sokal dislike besides truth relativity.
Technology may even be more fertile ground for these views than pure science. Those of us who have been involved in support and funding for technological projects see what the SSK people call self-fulfilling prophecies everywhere we have been. Consider an example from formal methods in computer and software development. A decade or so ago, the head of DARPA ISTO was rumored to have proclaimed "formal methods don't work" and lo, progress dwindled to a trickle in the US except for a couple of outstanding results. In contrast, the UK MoD wrote the use of state-of-the-art mathematical methods into relevant standards for computer-based system procurement (Def Stans 00-55 and 00-56) and lo, those developing safety-critical software in GB must comply, and Lockheed Martin must develop all the software for the C130J using a formal static analysis tool (SPARK), which development showed weaknesses in software developed according to the US civilian certification criterion DO-178B (for example, the proportion of significant errors in software developed according to DO-178's most stringent category, Level A, was indistinguishable from the proportion of errors in software developed according to the less stringent Level B). So development was proceeding according to two incompatible paradigms in two communities, which two communities overlapped significantly. Those two incompatible paradigms can't both be right, can they?
To interject an anecdote here, during my work in the US, I found a pervasive anathema even to simple math and formal reasoning in the US computer-based systems industry (an odd phenomenon, since graduate-level education in these fields requires at least some from everybody). In fact this partly led to my return to Europe, where math is a daily enterprise in technology, like eating lunch. I expected that the aversion to, respectively the facility with mathematics would show itself in a difference in the characteristics of software available. But this seems generally not to be the case, as far as I can tell. So, it could be argued that even these two partially contrary paradigms, lots o' math versus little math, did not seem to differ much in their effects. Nevertheless, very specific software such as that developed for the C130J is argued to be superior based on its use of some formal methods, and there seems to be justification for this view.
It seems to be generally accepted in computing technology that having the "best" product or using the "best" techniques accounts only for a small proportion of what happens in the market. And it is supposed that the market controls what gets developed and what not. Taken literally, this means that by far the larger share of the results is due to social factors, factors to do with the interaction between people, and thus better left to sociologists to explain. One may try to distinguish the situation of technology somewhat from that of science by observing that technology is more concerned with efficiency in practice, which is a highly social concept, rather than simply truth. But whether this brief observation can claim to be an explanation of the phenomenon is to me unclear.
I don't believe that facts such as 2+2=4 are so merely because of intersubjective agreement, because I can imagine a world in which everybody comes up with the answer 5 every time that they considered adding 2 to 2 (three year olds make such mistakes in patterns all the time - it would simply be a world in which everyone had the mental age of any 3 year old). But that does not make 2+2=5 and it is hard to believe that anyone should think coherently that it does, except a radical sceptic even further out than Descartes' doubter (can even Descartes' demon make P and not P both true at the same time?). So does that mean that I cannot regard facts as socially constructed?
In some sense, yes, but I think that interpretation misunderstands the thesis as proposed in SSK. SSK practitioners talk about "knowledge", taken to mean "shared institutionalized belief", and it is almost a platitude to say that shared institutionalized belief is socially constructed. Ergo, in this terminology, knowledge is socially constructed.
It seems that scientific theories can indeed sometimes be considered to be generated by social processes (namely, "shared institutionalized belief") similar to those that generate myths or "narrations". Consider my work reported above. Most of what I did is read the work of others, and through that reading came to form my ideas. Anyone studying what I did, or studying the development of a concept or theory through the work of the involved scientists needs a term for this iterated communication, and a term for its results, its "shared institutionalised meaning". It is like some kind of socially developed story. Why not call it a "narration"?
Consider an example. I looked at the evidence for the fact that single event effects in silicon devices (in which individual bits sporadically get stuck or flipped) at altitude (30,000 - 50,000+ ft) are caused by atmospheric neutrons, which, if you read the literature on it, is "generally accepted". Now, there are other possible explanations of SEE, and I wanted to know how they were ruled out, so I traced all the references to this "generally accepted" fact back. They either petered out, or terminated in one paper narrating some somewhat sparse tests performed at Los Alamos in the neutron irradiation facility in which DRAMs were partially exposed to neutron irradiation. Yes, there seemed to be an effect, but there were also two other paradoxical effects, orders of magnitude greater than the effect later claimed as "established". One was that the DRAMs from one manufacturer showed effects that were orders of magnitude greater that those from others; and the other was that in the DRAMs from this manufacturer, the number of bits flipped one way was orders of magnitude greater than the number of bits flipped the other way. Which is a phenomenon crying out for explanation, since the theory of neutron-induced effects predicts that the effects should be symmetrical. Neither of these two effects was explained in the paper. An observer used to applying any general theory of confirmation would note that the largest effect (by orders of magnitude) was one that contradicted the predictive consequences of the hypothesis under test (that SEEs were caused by the results of elastic collisions between silicon atoms and neutrons, and therefore that the effects should be symmetrical in which way bits were flipped) and therefore it seemed that the hypothesis was prima facie disconfirmed by the experiments. But yet the experimenters considered the hypothesis confirmed, without bothering to explain the anomaly, and this became the referential basis (in so far as there is one in the literature) for the claim that SEEs at altitude are due to atmospheric neutrons. This seems to me a clear example of a scientific theory being constructed by narrative processes, and a scientific "fact" being established by conducing its repetition in the literature. Apparently, this is a relatively common phenomenon, according to physicist colleagues with whom I discussed it. It is regarded as uncommon, though, in areas which attract a lot of attention.
There are also scientific theories whose predictions, while regarded as both true and consequential for human beings, concern effects which are so small in comparison with the influence of the background that they cannot, in principle, be reliably observed. One example is exposure to radioactivity and EM radiation of very short wavelengths associated with it. Estimates of the effects of low doses on humans are derived, as they have been for 50 years, by extrapolation of the measured effects of very high doses on the midcentury residents of Hiroshima and Nagasaki. The effects have a latency of a decade or more: effects are still being seen in survivors over 50 years later. They are mostly formation of organ and skin cancers, as well as other effects such as cataract formation. However, because not all potential lifestyle and environmental parameters have or can be recorded and measured, the data consist largely of cancer mortality statistics. Such statistics are confounded by unrecorded known correlants, such as whether subjects smoked or not and how much. No definitive statistical effect on health has been seen below a cumulative exposure of about 100 milliSieverts, which is more than most designated "radiation workers" such as airline pilots and atomic power plant workers receive in a lifetime's exposure. Recording exposure to the Hiroshima-Nagasaki bombs is not trivial - it affected survivors according to how much and what kind of material protected them, for example, and that is hard to assess except anecdotally. Measurements of radiation exposure of groups of "radiation workers" is through small badges, which are by no means as accurate as the significantly larger detection devices built by specialists for high-energy particle physics experimental research establishments. Cosmic radiation at altitude is also measured by smaller, less accurate instruments sent up in balloons from specific locations; the exposure of air crew to cosmic radiation is not identical to the background, since some of it is attentuated an unmeasured amount by the skin of the airplane and whatever extra material lies between outside and crewmember. Measurements inside the hull suffer from inaccuracy, because only lightweight devices may be used, and the spectrum is broad, and types of radiation varied. Offspring of aircrew who fly whle pregnant may exceed the 100 mSv exposure level, but they do not constitute any level of occurrence comparable even with the standard deviation of illness occurrence. Such low-dose effects, in short, are in principle immeasurable. Any effects inferred by extrapolation from high-dosage subject experience lie within the normal statistical deviation expected within the data gathered over years from the population at large.
While these effects are in principle unobservable, that does not mean they are not there. So an extrapolation is made from high-dose effects, and advice based on this extrapolation is given to potentially exposed humans, for example in US FAA Advisory Circular FAA AC 120-52, Radiation Exposure of Air Carrier Crewmembers, March 5, 1990. This AC, similarly to advice given out on the matter by the Health Physics Society and other official bodies, treats the matter of low-dose radiation effects on humans as fact. Whereas it is just a uniformly accepted guess, any evidence for which is buried in the statistical noise. It seems appropriate, even as a scientist who believes in facts the way Dr. Johnson believed in rocks, to say that there is and can be no fact of the matter here. But it is presented by some as if there were a fact of the matter. I deem it appropriate to call this a "myth" or "narration", for these words refer to socially constructed explanations, as long as it is not thereby inferred that the putative "facts" are thereby judged false. For these they cannot be judged false. They are consistent with all evidence, but nothing more and also nothing less. There is a lot more that would have to be known for them to become scientific fact, but they are treated as fact by many if not most scientists and technologists.
MacKenzie has observed a similar phenomenon in the determination of the "error budget" associated with targeting errors of long-range missiles in his book Inventing Accuracy (MIT Press, 1990). This determination in the face of principled immeasurability is by no means an isolated scientific phenomenon.
It may be objected that not all science is like that. Maybe so, but this "fact construction" does appear to be a common procedure amongst scientists. Further, I doubt that these cases appear sufficiently distinct in terms of the behavior of their participants from other cases in which the facts appear determinable (measuring the mass of an electron, for example). If one word is chosen to refer to the behavior of participants in a scientific enterprise such as determining which low doses of radiation are risky, whether atmospheric neutrons are largely to blame for SEEs at altitude, and what the charge on an electron is, then "myth" or "narration" would seem to be as good as any, as long as they are understood as technical terms for a particular social construction.
Consider now whether scientific debates are resolved by "rhetoric" or by "enlisting allies". First, let us consider rhetoric. I use the word "rhetoric" to refer to those features of any piece of argumentative writing which contribute (positively or negatively) to its persuasive force. This use is common amongst those who study this phenomenon. However, many people use the word to refer only to persuasive means which are regarded as logically illegitimate. I suppose it would be acceptable to Bricmont and Sokal to claim that scientific debates are resolved through considered presentation of explanatory arguments, and thus by "rhetoric" in my sense of the term. In that sense, it seems incontrovertible that scientific debates are resolved by rhetoric.
One may go further. There are very few scientific papers in which conclusions are established by rigorous, deductively valid and complete arguments. In fact, there are only two or three which I know, although I work in the field. I constructed one (even this was only a close approximation to one: I assumed that any logical truth could be used as an axiom) to see what it would take, using the Temporal Logic of Actions (TLA) of Leslie Lamport. Another, longer such proof in TLA was written by my student Dirk Henkel, who performed a rigorous verification that an algorithm which everybody already knew worked was in fact implemented by a specific algorithmic mechanism. The proof consisted of 115 pages of pure formal logic, and assumed the tautologies of classical propositional logic and the logical truths of classical predicate logic only. Lamport himself advocates rigorous proofs of algorithms to a level of detail unaccustomed even in mathematics, because the kinds of algorithms which he verifies are prone to mistakes of detail. However, he advocates terminating the steps of a proof at a much higher mathematical level than Henkel or I approached. The kinds of proofs performed by Lamport in TLA and Henkel in TLA are the most rigorous deductive arguments I know, but even they assume that the reader takes some principles (tautologies of propositional logic, theorems of predicate logic, not all of which are obvious) on faith. Scientific papers, even other correctness proofs of algorithms, do not begin to approach such deductive completeness. I would like to consider what this may mean.
A correct deductive argument is one in which every step follows from previous steps by deductively valid inference. Only deductively valid inferences can guarantee that their conclusions are true if their premises are. If an argument is not rigorous, some element of the truth of the conclusion is being taken on faith by the reader. Further, if the conclusion is to be established as true, all the premises of the argument, those principles which are assumed in order for the rigorous argument to be presented, must be taken to be true. I define a deductively rigorous argument as a correct deductive argument whose deductive steps are those of the proof rules of an accepted deductive system. According to this terminology, Lamport's TLA proofs are correct deductive arguments, and Henkel's are (almost) deductively rigorous arguments. The point of deductive rigor is that not only are the arguments correct deductive arguments, but they may be seen to be correct by checking each step simply and mechanically, whereas a correct, but not necessarily rigorous, deductive argument may use derived proof rules that are not obviously seen to be correct unless one has significant additional knowledge.
Most, indeed practically all, scientific papers do not consist of deductively rigorous arguments. Were they to do so, they would be unreadable (as Dirk Henkel's is, except by experts with a lot of patience). Hence the persuasive nature of a scientific paper largely comes from somewhere other than the deductive rigor (in my sense) of its argumentation. And any other argumentation styles besides deductive validity are fallible. Thus good scientific papers consist largely of fallible but persuasive argumentation, not of deductively rigorous (in my sense) arguments. In the majority of papers, it is even doubtful whether the arguments therein are correct deductive arguments. And to repeat, one can argue that only a deductively rigorous argument can without further ado be accepted as establishing its conclusion (and additionally its premises must all be accepted as true for it to do so). Any other than a deductively rigorous argument can rightly and literally be called a stepwise fallible attempt at persuasion. I think it is appropriate to use the word "rhetoric" for a stepwise fallible attempt at persuasion.
So much for rhetoric. What about "enlisting allies"? Are scientific issues resolved by rhetoric of the sort which consists laregely in enlisting allies, without any serious attempt to engage the argumentation of the supposed adversary? Most certainly. It is pervasive. But now we enter the realm of what most scientists consider anecdotes and prefer not to talk about in public, like the machinations of our colleagues on faculty committees. But it happens, and it is legitimate to recount incidents. Let me recount three such recent experiences of mine.
First, I wrote commentaries on two public documents concerning accidents with a US military development aircraft. One was the transcript of the public briefing of the inquiry board into the causes of an accident; the other was the report of a specially-convened panel into the development of the aircraft, which had been generally acknowledged to be troubled. I concluded that the inquiry board briefers had given a misleading characterisation of the cause of the accident, and confirmed my reasoning by interviewing a panel member, who worked at a major US university, by telephone. A colleague of his, who is a frequent discussant on mailing lists which I frequent, told me I was publishing mistaken factual claims, and that I had unethically misled the panel member. Furthermore, my colleague informed me that he was privy to the truth but could not talk about it for professional reasons. I was concerned about the claims that I had acted unethically, because they were in fact mistaken. I asked two other US colleagues what to do. One said it was a known behavior of my critic. The other said that since my critic is an professor at a renowned institution, I had not only to assert that the claims of unethical conduct were mistaken, but persuade my accusor of the validity of my arguments (even though prima facie I had failed at that already). Neither of my advisors commented on the validity of my analysis of the public pronouncements on the aircraft. Their commentary was 100 per cent social, and for good reason: the arguments presented by my critic were almost 100% social, with the sole exception of the assertion that my analysis was wrong, for which no ground was given. There was no attempt to go through my analysis and the source documents on which it was based and show where the analysis was incorrect and where correct.
Second, a colleague publically mischaracterised a method which I have developed for causal analysis of system behavior, on a mailing list containing experts with an interest in my methods. In response, I went back to the published classification used to (mis)characterise my method, and pointed out how my method fit into a number of different other categories as well; however, the classification was supposedly based on exclusive categories. So there must be something wrong either with the categories or with the way they were used in classification. I also expressed my wish (and expectation) that my colleague would engage in a substantial debate. My colleague dismissed the request for discussion by claiming to be too busy dealing with "real" problems to engage in an "academic pissing match". That was the end of the public discussion. The rhetoric used by my colleague was designed to enlist allies (as most public expressions of outrage are), and not to engage in a serious discussion of the classification of my method. This brief description should suffices to show that anyone's opinions on my work and that of my colleague were not intended to be formed by careful analysis of our technical work, but rather by a social process of some sort which involved little technical detail.
This kind of interaction, based primarily on forms of status and alliance, happens frequently to almost everyone I know, and is as pervasive a form of rhetoric as technical argumentation, if not more so. How can we pretend that social manoeuvrings are not a major part of the persuasive techniques we use to (try to) prevail? I don't think we can.
Third example. I recently engaged in a debate on a mailing list composed of pilots, engineers, regulators and others concerned with use of computerised flight management systems on commercial aircraft. An engineer claimed he had built systems to a failure rate of 1 in a billion hours, and his experience confirmed that indeed these systems were so accurate. This specific failure rate is a certification criterion for safety-critical systems and subsystems aboard commercial aircraft, and has been so for decades. Such claims, however, while required for certification, cannot feasibly be confirmed through testing or experience with the systems, unless one is in advance already convinced by other means that the system is that reliable. This fact is established by theorems in the literature. So even were a system to be so reliable, one cannot know it. This phenomenon is recognised by UK MoD Def Stan 00-56, for example. Further, a colleague in the UK, who is an acclaimed safety-critical systems software entrepreneur, believes that no one can show him a 1-in-a-billion-hours argument of any sort that he couldn't rubbish. (I am inclined to make the same claim, but I have not seen as many such arguments as he has.)
My arguments consisted of citing the literature and showing how it contradicted what the engineer claimed was his practice. His arguments consisted of explaining the way he works (as do others following this methodology), and claiming I was unfamiliar with the details of the working. He also made various claims concerning my motivation for engaging him in discussion and not accepting his statements. His arguments seemed to me to be little more than attempts to enlist allies from amongst his community (which he succeeded in doing to some extent). On the other hand, I certainly admit to hoping that my arguments would succeed in enlisting allies on my side. So that is what we were both doing, for better or worse. How can one deny it?
Many if not most of my UK colleagues would make the same arguments as I did, based on the same understanding of processes concerning estimation of failure rates. My peers in the US know the science, and accept the reasoning, but some or even many of them defend the industrial methods, even though using those methods requires one to make claims that contradict the science. And, indeed, those methods do appear to work! Airplanes do not, in general, fall out of the sky. We must conclude that the reasons they work cannot be identical with the claims those methods themselves make as to why they work, on pain of contradiction. So the claims those successful methods make are indeed "myths", in the strong sense that some of them must be untrue. But they are used. They work. They continue to work. And their practitioners continue to advocate the demonstrably false parts. What else are we to make of this phenomenon, other than to characterise it as a "myth" which is nevertheless successful?
I think these examples show that it is even possible to be a naive realist about science, and still consider much technological knowledge (in the sense of shared institutionalised beliefs) socially constructed (the one-in-a-billion process certainly cannot be described as veridical by a naive realist); that the theories or accepted practices have similar characteristics to myths or other communicative developments (same example); that debates are often resolved by rhetoric in its perjorative sense or by enlisting allies (all three examples). As to truth, I have argued for an interpretation of SP in which it proposes no truth theory (as opposed to, say, a truth-relativity theory), but rather ignores the issue. I don't see that there is so much very wrong with the presumptions that Bricmont and Sokal consider equivalent to radical scepticism, although I have great reservations about radical scepticism, as do most people who look before they cross the road or write papers that they expect might be read.
At Friedman's meta-paradigm level lie the Euclidean geometry of Newton or the Riemannian geometry of Einstein. These clusters of mathematical knowledge are supposed to have a validity independent of the fate of Newtonian or relativistic mechanics, as indeed all mathematicians believe they do. There is not supposed to be a way these intellectual constructions can be refuted, except by exhibiting internal contradictions in them (it is in some sense always thinkable that there should be such that no one has yet discovered, but no competent mathematician believes it. It has been supposed at times that formal arithmetic may be inconsistent, in particular since the work of Gödel, indeed so-called intuitionist mathematicians accept only parts of classical arithmetic, but there is little doubt that, were arithmetic to be found inconsistent in some way, revised principles of arithmetic would be devised that preserved at least the arithmetic of shopkeepers). At the level of "properly empirical" laws which face squarely the "tribunal of experience", there is no such insulation from refutation. Thus the meta-framework and the empirical level must be distinct. There must then be principles explaining how the meta-framework is used to formulate the empirical law level: principles of translation that say how features of the meta-framework are to be interpreted in the physical universe such that the physical laws framed in their terms can be adjudged.
Friedman considers the Newtonian law of universal gravitation, and the Einsteinian gravitational field equations, as constituting the empirical level at which the respective theories of mechanics can be tested, say, against the observations of the perihelion of Mercury
At the level of connections, Friedman points out that the Einsteinian principle of equivalence "depicts the space-time trajectories of bodies affected only by gravitation as geodesics or straightest possible paths in a variably-curved space-time geometry", just as the Newtonian laws of motion "depict the trajectories of bodies affected by no forces at all as geodesics or straightest possible paths in a flat or Euclidean space-time geometry."
Einstein's field equations describe the variations in curvature of space-time geometry as a function of the distribution of mass and energy. This is only possible against a background in which space-time can be taken to have such a geometry (which became available after Riemann but not before); that is, in which some empirically given phenomena are selected as counterparts of the fundamental geometrical notions, in particular geodesics. Einstein's principle of equivalence does this. Hence Friedman's identification of the principle of equivalence as the connecting layer between the meta-paradigm, the geometry, whose intellectual worth is irrespective of physical observations, and the conjectures which are most subject to confirmation or denial according to testing against physical phenomena. Similarly, the Newtonian laws of motion explain how the Euclidean geometry is interpreted in space and time so that the law of universal gravitation can be seen to fit or misfit our observations.
My point here is not to establish and explain the three-layer stratification of physical theory counterposed to an undifferentiated Quine-Duhemian holism, say, but to establish the idea of partial incommensurability of theories against a Kuhnian suggestion that theories are either commensurable or not, with no in between stages distinguishable. I believe Friedman has made a plausible attempt to distinguish levels of a physical theory which accords with practioners' understanding and indeed with physical practice. Whether one can disentangle the correspondence principles - those that interpret the geometry, say, in the realm of physical phenomena - from the empirical principles of a scientific theory is an issue which need not engage us. The relevant point is that both Newtonians and relativists can talk to each other, via a mutually-comprehensible explanation of the different geometries underlying the two theories. Although the theories are ultimately different, they are partially commensurable. One way in which one might argue for the tripartite separation is through the event of Cartan's reformulation of Newtonian mechanics using Riemannian geometry and the principle of equivalence. The Cartanian formulation of Newtonian theory is equivalent in its empirical consequences to Newtonian theory; the Cartanian formulation and the Einsteinian theory share the meta-paradigmatic background of Riemannian geometry as well as the translation principle of equivalence. Cartan demonstrated clearly that Newtonian theory and Einsteinian theory were more commensurable than had previously been thought. One may conclude that partial commensurability is a fact for scientific theories.
Not just for scientific theories, but also for pure principles of deductive reasoning. Intuitionist mathematicians differ from classical mathematicians on (amongst other things) the principles of logic, of valid deductive reasoning, which they accept and use. Work of Kleene, Heyting and others showed how, purely syntactically, the propositional deductive logics of the two schools are comparable. The semantics continue to differ under the comparison afforded by this work, but the principles accepted by intuitionists are syntactically a subset of those accepted by classical logicians. Further work by Dummett and Lemmon later showed how one may formulate deductive logics with a concept of necessity (so-called deictic modal logics), with identical semantic interpretational principles, in which the set of necessary principles of one logic are those of intuitionist propositional logic, and those of the other are those of classical propositional logic. So the two sets of deductive principles, although prima facie incomparable, turn out to have multiple modes of comparison, differing in the structural nearness which they can elicit from the divergent logics.
Peter B. Ladkin