The tentative nature of science

Ars Technica has an interesting post on how scientists themselves view the tentative nature of science. In ordinary language, a tentative conclusion is not to be preferred (the old "evolution is just a theory" canard), but in science it is in fact a virtue. Science's conclusions are meant to be tentative, but a paradox is that as its conclusions are used and extended, that tentative nature tends to evaporate over time.

John Timmer asked a number of scientists how they viewed their data, their models and the broad theoretical frameworks. He got differing answers from different fields. Physicists and biochemists treated data as more tentative than the models in the light of which they were interpreted. But genomics researchers treated their data as high quality, and a neurobiologist insisted that the data in the field were messy, but not tentative.

In the case of models - which are explanatory schemes for a given phenomenon, either rough conceptual schemes or formal mathematical descriptions - all agreed they were tentative. A common theme was that "[i]f confirmed data disagrees [with a model], then the scope of the model becomes more limited or the model must be modified or discarded entirely at some level." Interestingly and tellingly, biological models were regarded as restricted to the model organisms studied. I think this is due to the fact that generalisations in biology are based on historically contingent evolutionary sequences, and so the very next organism studied may have a divergent pattern from the previously studied ones.

What is most interesting is that theories, which form an intellectual framework for research and explanation, are regarded as much less tentative than models. For example (and this is my own example) natural selection was initially a rough model that became formalised by Fisher and Wright and subsequent theoreticians. It is now an established fact and broad theoretical framework in biology at all levels, although it is hardly the sole theory applied to all organisms. But even here, biology is something of an outlier, and theories are often abandoned or ignored, according to Timmer.

Engineering, which is the home of so many evolution deniers, is surprisingly also an outlier, in that major theoretical advances force major revisions every so often. Many engineering decisions are economic or political, though, and the chaos of development is masked by the final products and publications being so decisive.

Something Timmer alludes to but doesn't explicitly say, is that very nature of science is to make progress by rebuilding itself; to use Neurath's famous metaphor, much beloved of Popper, science is a boat being rebuilt at sea. Science learns by making conjectures, testing them, and occasionally abandoning them. Of course, it can't do this is every piece of data or theory or model is up for grabs, so the teantive nature of data and models is not universal, nor should it be. Information and ideas that have served well for a long time, contributing to ongoing active and fruitful research will tend, properly, to be accepted. I am reminded of Hull's interviews with evolutionary scientists:

Yet another ambiguity constantly crops up in our discussions of scientific theories. Are they hypotheses or facts? Can they be "proved"? Do scientists have the right to say that they "know" anything? While interviewing the scientists engaged in the controversies under investigation, I asked, "Do you think that science is provisional, that scientists have to be willing to reexamine any view that they hold if necessary?" All the scientists whom I interviewed responded affirmatively. Later, I asked, "Could evolutionary theory be false?" To this question I received three different answers. Most responded quite promptly that, no, it could not be false. Several opponents of the consensus then current responded that not only could it be false but also it was false. A very few smiled and asked me to clarify my question. "Yes, any scientific theory could be false in the abstract, but given the current state of knowledge, the basic axioms of evolutionary theory are likely to continue to stand up to investigation."

Philosophers tend to object to such conceptual plasticity. So do scientists -- when this plasticity works against them. Otherwise, they do not mind it at all. In fact, they get irritated when some pedant points it out. [David Hull Science as a Process, 1988, 7]

More like this

I've always thought it was ironic that one of science's greatest strengths -- its falsifibility, its readiness to drop past assertions when new information contradicts them -- is seen by Creationist as its greatest weakness. This suggests to me that these two completely different frames of thinking, so divergent and contradictory, can never be reconciled and the only resolution will be when one pushes the other out of the knowledge community.

When discussing the issue with non-scientists, I usually tell them that the sentence "science currently explains this phenomenon X through theory Y" is not a claim of the truth of Y, but "merely", that Y is the best we have so far for explaining X (with the "we" meaning humankind, not just scientists). They are taken aback by how weak this sounds; but after repeating the sentence a few times, stressing "the best we have", most intelligent people come to realize that, au contraire, the statement is as strong as it can get (since it means that any alternative explanation has been dropped -or put in second place- for some good reason). People grow up assuming that absolute truths exist, but except for a minority of fundamentalists (far smaller than the official number of "believers"), they eventually accept that this is not the case, at least in any practical sense.

By dileffante (not verified) on 14 Oct 2006 #permalink

Yes, I'd say one of the fundamental issues is just the difference in attitude toward "being right". On one side, there are people who defend their beliefs "to the death", because they need to "be right" past and present. On the other, stand those who are willing to let go of a belief, because they'd rather become "right" for the future -- even if that means changing their minds now.

By David Harmon (not verified) on 14 Oct 2006 #permalink

I wish the tentative nature of scientific knowledge was better understood by the general public, and have wondered how science educators could do a better job communicating this to students. Way back when I taught courses in Scientific Reasoning, students responded well to Popper's ideas about falsification. They almost always started with an image of science as the attempt to prove hypotheses, and actually seemed excited by the "reversal" proposed by Popper. Of course, falsification is not the whole story of science, and I tried to inculcate the view that both confirmation and falsification are important elements of scientific method. Overall, I tried to leave them with the idea that science is as much about raising and clarifying questions as about finding answers.

By bob koepp (not verified) on 14 Oct 2006 #permalink

Maybe it's the fact that I've read too much van Fraassen and model theory (mathematical logic), but I don't understand the distinction you're making between models and theories, for example when you say that "theories, which form an intellectual framework for research and explanation, are regarded as much less tentative than models." Isn't a theory just a concise and informative way of demarcating its class of models?

Second, I really don't think I'm understanding that second paragraph from Hull. "Philosophers tend to object to such conceptual plasticity." Wha? He's describing holism, after all, and holism's been prominent since Duhem first defended it more than a century ago. I was going to make a list, but it's already over eight people long, and that would just be tedious.

I'm aware that there are those who think theories are just collations of models, and in some respects that is true, but theories are, in my view, much more than that. Take, for instance, the Lotka-Volterra equation in ecology - as a model it represents the relations between eater and food, but to apply it one also needs to know what conditions have to be bracketed out, what sorts of relations obtain in a particular case (is the eater a generalist or specialist, can it turn to other food sources, and so on). All of this, which is the interpretation of the model, is external to it.

Similarly population genetics makes a host of simplifying assumptions about the independence of genes, what the relevant genes are, and so forth. These ancillary assumptions form part of a wider semantic and conceptual framework in which the practices of the discipline play a major role. This is the theory - a semantic structure that can be employed, and the formalisms of the models are applied within it.

Having said that, there is no simple class of conceptual frameworks that covers as all and only theories. Theories are historical entities that exist within a particular discipline and each discipline is sui generis to some degree, until we have a way of reducing the theories of one into the theories of another more general discipline or domain.

The philosophers Hull refers to are analytic philosophers of science, who need to see things in sharp and clear terms, when (as Hull proceeds to argue) science is anything but a clear and distinct set of conceptual commitments. I don't see what holism has to do with this, and anyway I think "holism" is a general term either of abuse or approbation depending on the perspective of the person using it.

Phil: on facts -

There's an analytic sense of fact that I agree (with both you and Quine) no longer applies, but a fact is something that is so deeply embedded in our cognitive structures that to deny it would cause irreparable damage to those structures. I cannot think that the orbit of the earth around the barycentre of the sun is not a fact. I cannot think that natural selection is not a fact. These things have been demonstrated and cohere with so much of our scientific conceptual scheme that to deny them, even if conceptually possible, is to reject knowledge. I'm not appealing to Protokolsätze or to sense impressions, but observations of things and their continuing successful integration into our corpus of knowledge qualifies as as much of factuality as we ever need, IMO.

"Interestingly and tellingly, biological models were regarded as restricted to the model organisms studied."

This is a very odd statement, given the prominent role played by model organism approaches in biomedical research, and the amount of money spent by the NIH on model organism research. The evolutionary relationship between all organisms means that biological models developed using one organism are frequently applicable to others. Of course, it is necessary to determine to what extent this is true in a given instance.

I try to inculcate into my trainees the following attitude towards models and hypotheses: We don't care at all whether our models and hypotheses turn out to be "right" or "wrong". All we care about is that they lead to interesting new experiments.

By PhysioProf (not verified) on 14 Oct 2006 #permalink

The use of murine, bovine, piscine and other animal models in, say, medical research is inductively generalisable just to the extent that they are phylogenetically related to humans. But a particular model - for instance Drosophila - will have its own uniquely evolved traits, and to do the generalisation one has to be sure that the traits under study are either synapomorphies or plesiomorphies of the clade that contains both. I gather that some genetics done on Drosophila are in fact unique to that clade, but were previously thought of as general. Likewise the initial genetics done on transcription in bacteria turn out not to be universally true in eukaryotes, or even in any.

There's no alternative to model organism research - it's the only entré to studying organisms at all, but we have a lot to learn about both humans and other organisms not studied. Consequently, there are no laws as such in biology (which aren't laws of chemistry or physics) because the states and processes studied are historical and contingent upon the evolution of the clades concerned. As Hull said, echoing Cicero on philosophers, there's nothing so odd or absurd that some organism somewhere doesn't exhibit it.

Thanks for clearing up your usage of `theory' -- I think there really is just a difference of definition here, nothing substantial.

By `holism', I mean the view that answers `Yes' to Hull's question "Do you think that science is provisional, that scientists have to be willing to reexamine any view that they hold if necessary?" There's probably a bit more common ground than *just* that between Duhem, Neurath, Quine, and other philosophers readily identified as holists, and you're certainly right that `holist' can be a vague term of abuse or approbation. But if we set aside trying to define `holism', then I think my basic point still stands: prima facie, Hull seems to be saying that philosophers `tend to' answer `No', despite the prominence of philosophers who would answer `Yes'.

Of course, there's that `tend to', and the relationship between `conceptual plasticity' and the way someone answers that question might be tricky. And seeking a precise exegesis based on two paragraphs is a stupid undertaking. So I'll just say that the little cartoon question mark popped on when I read the quotation, I'm going to be thinking about it when/if I ever happen to read this book.

I think a mild paradox is at work here. Precisely because the sciences have arrived at conclusions which, if not indesputable, are at least vastly better grounded than the notions promoted by theologists and specualtive philosophers, scientists get a rhetorical advantage by insisting that their results are provisional. After all, if heliocentrism or evolution is merely provisional, what does that make of the bulk of received wisdom, whose basic warrant is that it has been believed by a lot of people for a long time?

I don't think there's a single answer to questions about the status of scientific results because the right answers differ depending on who's asking 'em and why. For a logician, scientific results are, to use a prehistoric terminology, synthetic statements and therefore in principle falsifiable. Except that if you're a philosopher of science, you notice that some apparently synthetic statements are treated as analytic in practice. Meanwhile, as I suggested above, a practicing scientist has every reason to be scrupulous about insisting that all work is tentative even if it turns out that believing six known things before breakfast is at least as hard as believing six impossible things. And as a consumer of scientific information, I have much better reasons to believe statements such as "the Earth is of very great antiquity" than statements such as "my name is Jim." For us civilians, operating within the normal rules for evaluating hearsay, the core scientific results are knowledge if anything is knowledge.

I would still follow Quine and dismantle the disctiction between analytic and synthetic statements. Even as a logician, I would say the rules of logic are synthetic as well. I haven't seen any good argument against this. When we talk of the 'solididy' of some statements over others, other considerations -- whatever they are -- come into play.

It seems to me that Quine doesn't and indeed can't dispense with the distinction between analytic and synthetic precisely because he needs the distinction in order to make a point about how the logical status of statements depends on how a statement is being used relative to the state of play of actual sciences. Am I missing something? That would hardly be surprising since I haven't reread Quine for maybe a quarter of a century!

By the way, Phil Thrift writes: "Even as a logician, I would say the rules of logic are synthetic as well." Does that claim imply that at least some statements are analytic, even if the rules of logic are not? And if so, are you dispensing with the analytic/synthetic distinction or simply applying it differently? Or maybe the analytic/synthetic distinction is like the ladder in the Tractatus. Once you're on the roof, you can throw it away.