More Recent Comments

Thursday, July 12, 2007

Is There a Correct Way to Do Science?

 
I think there are incorrect ways of doing science and I think that much of today's scientific literature is an example of bad science. This isn't news. Back in 1963 Peter Medawar felt the same way and he expressed this viewpoint on a BBC radio show. The transcript, Is the scientific paper a fraud, is published in Medawar's collection of essays called "The Strange Case of the Spotted Mice" (Medawar, 1996). (The title of the book refers to a scientific fraud in immunology that Medawar uncovered in the early 1970's.)

Medawar says,
... the scientific paper may be a fraud because it misrepresents the processes of thought that accompanied or gave rise to the work that is described in the paper.
Janet Stemwedel discuses this paper on her blog Adventures in Ethics and Science [Why does Medawar hate the scientific paper?]. Janet uses the paper as a way of introducing some key concepts in epistemology—loosely defined as "the investigation of the origin, nature, methods, and limits of knowldege." (Burr and Goldinger, 1980). Read Janet's blog and the comments in order to see this perspective.

I want to take the discussion in another direction. Medawar's complaint is that the scientific paper distorts the real process of science by misrepresenting the steps that are actually followed in a scientific investigation. In particular, the typical paper suggests that induction is the main mechanism of scientific discovery. Here's how Medawar describes the naive scientific method,
What induction implies in its cruder form is roughly speaking this: scientific discovery, or the formulation of scientific theory, starts with the unvarnished and unembroidered evidence of the senses. It starts with simple observation—simple, unbiased, unprejudiced, naïve, or innocent observation—and out of this sensory evidence, embroidered in the form of simple propositions or declarations of fact, generalizations will grow up and take shape, almost as if some process of crystallization or condensation were taking place. Out of a disorderly array of facts, an orderly theory, an orderly general statement, will somehow emerge.
Now, nobody has ever accused Peter Medawar of being stupid so before you start to quibble about this sort of generalization, be aware that Medawar does not apply it to all of science and every scientific paper. He's talking about common, but not exclusive, practice.

One of the reasons why pure inductive reasoning is misleading is because we never start an investigation with a clean slate.
... the starting point of induction, naïve observation, innocent observation, is a mere philosophic fiction. There is no such thing as unprejudiced observation. Every act of observation we make is biased. What we see or otherwise sense is a function of what we have seen or sensed n the past.
This seems like something that's so obvious that it hardly deserves mentioning. But it does deserve mentioning. Medawar was right to have brought it out into the open and it's something we always need to keep in mind.

Now we get to a famous quotation from Medawar's talk. Janet Stemwedel discusses it in the context of Popper and falsification—conveniently ignoring Kuhn, who is far more relevant—but I want to use it to talk about adaptationism.
We wish to question a deeply engrained habit of thinking among students of evolution. We call it the adaptationist programme,
or the Panglossian paradigm.

Gould and Lewontin (1979)
What is wrong with the traditional form of the scientific paper is simply this: that all scientific work of an experimental or exploratory character starts with some expectation about the outcome of the inquiry. This expectation one starts with, this hypothesis one formulates, provides the initiative and incentive for the inquiry and governs its actual form. It is in the light of this expectation that some observations are held relevant and others not; that some methods are chosen, others discarded; that some experiments are done rather than others. It is only in the light of this prior expectation that the activities the scientist reports in his scientific papers really have any meaning at all.
The debate between the adaptationists and the pluralists is often dismissed—usually by the adaptationists—as mere quibbling about scientific data. After all, they argue, all we need to do is collect data on each characteristic under discussion in order to resolve the question; is it an adaptation or an accident?

They are wrong. This is not a debate about facts at all. It's a debate about the the prior expectations that one has before any data is collected. Criticism of the adaptationist program focuses on the state of mind that researchers posses when they are designing experiments and formulating hypotheses to test. In other words, it's what Medawar refers to when he says that scientists start with "some expectation of the outcome of the enquiry." It's their bias or prejudice that's being questioned and not just whether the horns of an Indian rhinoceros are adaptations or not [Visible Mutations and Evolution by Natural Selection].

One of the main arguments of the Intelligent Design Creationists is that modern scientists adopt a naturalist assumption whenever they do science. This rules out creationism by definition. Some scientists attempt to deny that scientists have any pre-conceived notions at all. They promote the myth of pure inductive reasoning driven only by facts and evidence. It would be wise to drop this form of argument since it is obviously false. Scientists do have prejudices and biases. We need to recognize them and try to deal with them. The essence of skepticism is to always question your assumptions and discard them if they cease to become useful. That's what distinguishes the scientist from the preacher.


Burr, John, R. and Goldinger, Milton (1980) Philosophy and Contemporary Issues Macmillan Publishing Co., New York

Gould, S.J. and Lewontin, R.C. (1979) The Spandrels of San Marco and the Panglossian paradigm: a critique of the adaptationist programme. Proc. R. Soc. Lond. B 205:581-598.

P. B. Medawar (1996) "Is the Scientific Paper Fradulent?" in THE STRANGE CASE OF THE SPOTTED MICE, Oxford University Press, Oxford.

3 comments :

Mike Haubrich, FCD said...

Sorry, I am not a scientist, but isn't the purpose of peer review to spot and correct incidences in which preconceptions have interfered with the conclusion?

I would say that often we don't recognize our preconceptions and need external feedback to uncover faults in design or conclusion which even the must erudite skeptic misses in his or her own work.

There is a saying in writing "You shouldn't rely on yourself as an editor."

Torbjörn Larsson said...
This comment has been removed by the author.
Torbjörn Larsson said...

I think you may misread Janet's post. She describes Medawar as discussing in Popperian terms as far as I can see. If you go back in here series of basic posts you will find that Janet isn't too hot on Popper IIRC.

Naive inductionism is one of my pet peeves, since it forgets that we do form working hypotheses, have standardized tests, and it is an iterative process that forms mutually interdependent networks of hyotheses, i.e. theories and networks of theories. Naive popperianism on hypotheses isn't enough either, but at least it can be easily extended to incorporate all of the above by testing on theories. (Albeit it seems Popper himself dismissed induction as important in the process.)

What I understand as the problem is when Popper tried to define what science is and how the scientific project works. I doubt Kuhn does it any better, but I hear that we get a distorted view of his thinking outside his books.

On the preconceptions, I think it may be a philosophical mistake. Since as you say we can judge the process by the result we can (a la Popper) disregard what assumptions we use exactly, I think. We can further view them as tested while testing results and so not being a priori assumptions in the strict sense.

But of course the result will somewhat be affected by working assumptions and available methods - the rate with which we work at certainly is. Perhaps there is an assumption of that ad hocs doesn't form the major part of theories and will eventually be replaced later here. (But in that case it is a small working assumption, right? :-)

Btw, when you discuss this I get the vision of induction as hypotheses drifting with the data and popperian testing (falsifiability) as selection on the hypotheses. :-P Seems we can use both.