Sunday, January 17, 2010

Scientific Methods

by Richard Crews
There is no such thing as "The Scientific Method"; that is a myth. It is a good myth, in a way--it is useful and reassuring: it is useful to those who popularize science, and reassuring to all the rest of us who "use" science every day. But it is a myth nonetheless.

The Wikipedia says that the term "scientific method" refers to "a body of techniques" that are used "for investigating phenomena [and] acquiring new knowledge." A method is called "scientific" if it is "based on gathering observable, empirical, and measurable evidence" and the thinking and evaluating processes used are "subject to . . . principles of reasoning."

That's pretty open ended--pretty loose--when you think about it. It is hardly the simple recipe we know and love (and commonly think is all there is to "the scientific method")--

(1) form a hypothesis (put it in such a form that we can state a "null hypothesis," that is, an alternate statement that says the variations that the hypothesis purports to explain might just be due to chance)

(2) design an experiment that will test this (in other words, that can possibly "defeat" the null hypothesis)

(3) do the experiment; make the observations

(4) show by statistical analysis that the null hypothesis is probably false--so the original hypothesis is possibly true.

This neat, four-step process--which many (sophisticated) people think is what constitutes "the scientific method"--is satisfying, I guess, when you can do it, but it often does not apply--it is entirely impossible in some branches of science.

In fact, there are five different kinds of "scientific methods" that are used in different scientific fields:

(1) Empirical Experiments--this is the approach described in the recipe above; it can often be used in "hard" sciences such as physics and chemistry, sometimes in biology--but never in such fields as archeology, anthropology, astronomy, and many others.

(2) Questionnaires--this approach is used in the social sciences (such as anthropology or epidemiology). One finds a group of people (such as fat people) and has them fill out a questionnaire about how they got along with their mothers and fathers when they were little. The same questionnaire is also given to a "control" group, that is another group of people of the same sex, age, cultural background, etc. And then one sees (through statistical analysis) if there are patterns of differences between the responses of the two groups that help "explain" why some people are fat.

(3) Natural Observations--this method is used, for example, in astronomy where there is no possible way to experiment (that is, to change the input conditions to see what happens). It is interesting that the next generation of telescopes now being deployed will make observations so deep in time and space that the data they measure will be a "one-off," they can never be expanded or heightened; once the deep data on the origins of our Universe have been observed, there will never be comparable observations of another Universe. It has therefore been proposed that substantial portions of these new data be withheld from theoreticians for several years so that they can form hypotheses, and then have a chance to have these hypotheses "tested" against new batches of the primary data.

(4) Thought Experiments--imagine, said Einstein a hundred years ago, that you are in a dark elevator falling at a rapid rate out in inter-stellar space, and there is a beam of light shining in through a hole in one wall of the elevator. Of course you can't really conduct this experiment and make the observations that Einstein described, but you can imagine what you might see. Einstein used such "thought experiments" to demonstrate the implications of his theory of relativity.

(5) Simulations--over the past few decades as computers have gotten more and more powerful, it has become possible to simulate vast patterns of weather or climate changes, of stars evolving or exploding, of clouds moving, or of flocks of birds obeying simple rules of flight, etc. One programs a computer with the variables one wants to study, and then runs the simulation to see how it comes out. In a sense these are thought experiments, but they are much more complex than a human mind could handle.

So "scientific methods" vary from one field of science to another, and one scientific question to another. And how these methods are applied is clearly dependent on such factors as the cleverness, personal background, and cultural setting of the scientists who use them.