Sunday, June 26, 2011

Scientific Method in Practice (Pt. 5)

In this series of posts, I'm re-reading Hugh G. Gauch, Jr.'s philosophy of science textbook Scientific Method in Practice (Google Books).

[Series Index]

Recent History

Just as Part 2 of this series covered the four 'bold claims' of rationality, truth, objectivity, and realism, this post will focus on 'four recent woes' which 20th century philosophers have advanced: elusive truth, underdetermined theory, incommensurable paradigms, and redesigned goals.

Elusive Truth

I may be the only person whose first philosophy book was Karl Popper's Realism and the Aim of Science. His falsification criterion of separating science from pseudoscience appealed to me then, and I still frequently hear it invoked in discussions about science. Essentially, a theory or hypothesis isn't properly scientific if it can be used to explain just any data whatsoever; it must risk disconfirmation by implying what can't happen if it's true. This is what it means to be 'falsifiable.'

In Popper's day, the theories of Marx and Freud 'made sense' to huge numbers of people but seemed capable of interpreting anything as a validation. Meanwhile, Einstein's theory of relativity was very counter-intuitive, yet it made risky predictions that kept failing to fail! This limit (or 'demarcation') on proper science has been taken to heart in contemporary debates on whether evolution, intelligent design, climate change, etc. are scientific views.1

But wait, where does the 'elusive truth' woe come in? From overcompensating to the point of denying any place for induction in scientific method. Notice how falsifying a hypothesis is a deductive affair: if P then not Q; Q; therefore not P. In Popper's view, science can only proceed in this negative fashion of eliminating false theories, without any way to add positive weight to the likelihood a theory is true.
Scientific theories can never be 'justified', or verified. But in spite of this, a hypothesis A can under certain circumstances achieve more than a hypothesis B—perhaps because B is contradicted by certain results of observations, and therefore 'falsified' by them, whereas A is not falsified; or perhaps because a greater number of predictions can be derived with the help of A than with the help of B. The best we can say of a hypothesis is that up to now it has been able to show its worth, and that it has been more successful than other hypotheses although, in principle, it can never be justified, verified, or even shown to be probable. This appraisal of the hypothesis relies solely upon deductive consequences (predictions) which may be drawn from the hypothesis. There is no need even to mention induction.2
As a toy example, let's say I have a small bag of marbles. After shaking the bag, drawing out one marble, seeing it's red, and putting it back in the bag, I conjecture that all the marbles are red. If the next marble I draw out is blue, my conjecture is refuted. But what if I draw out a red marble the next 10 times? The next 1,000 times? The next 1,000,000 times?

From a purely deductive point of view, we are no more justified in thinking the bag contains all red marbles after a million trials than after one. The conjecture that the bag is a half-and-half mix of red and blue marbles is equally justified. Apply this to scientific method in general and we must give up our confidence that well-tested theories are close to the truth.

Underdetermined Theory

Two points on an XY coordinate graph. Which equation fits the points? If you answered 'infinitely many equations can fit those points,' you win! This holds true as the number of points increases. As a computer graphics textbook puts it:
Given n data points that are numbered P1 through Pn, there are infinitely many curves that pass through all the points in order of their numbers, but the eye often tends to trace one imaginary smooth curve through the points, especially if the points are arranged in a familiar pattern. It is therefore useful to have an algorithm that does the same.3
I hope the analogy to observations (points) and scientific theories (equations) is clear. As human beings, we may tend to prefer a certain style of theory 'fit' to our observations, but there are an infinite number of other possible theories which can fit the data. In other words, the data does not uniquely pick out (determine) a theory. We'll come back to this idea in a moment.

In his book Patterns of Discovery: An Inquiry into the Conceptual Foundations of Science, Norwood Russell Hanson argued that observation is dependent on theory (i.e. observation is theory-laden).4 This threatens to systematically undermine the distinction between theory and observation. And if that happens, then the problem of underdetermined theory produces the problem of underdetermined observation.

What does this mean for falsification? It is reduced to a mere coherency check between the theories underlying an observation and the conjectured theory. A contrary observation cannot prove a theory false, only inconsistent with some other theories.
So the first woe cited earlier was that science could not verify truths. Now this second woe is that science cannot falsify errors either. Science cannot declare any theory either true or false!5
Incommensurable Paradigms

In The Structure of Scientific Revolutions, Thomas S. Kuhn outlined a new model for the way science changes over time. Instead of seeing science as steady progress occasionally hastened or corrected by geniuses, he identified two radically different modes.

During periods of normal science, the scientific community tacitly agrees to a set of assumptions and works on a set of puzzles they currently find interesting. They make steady progress on these puzzles using their shared assumptions. This combination of the (current) way science is done and the (current) objects of scientific inquiry constitute a paradigm.

Periods of revolutionary science begin when new assumptions or new puzzles disrupt the old fit between the two. There is no rational way to say one set of assumptions is more correct than another, or that one set of puzzles is more important than other. When science finishes shifting to a new paradigm, it's a matter of changing to match circumstances not improving against a common measure.

The paradigms of two periods of normal science are incommensurable, which means not-commonly-measurable. So why do we tend to think Einstein's physics was an improvement over Newton's physics, and Newton's physics an improvement over Aristotle's physics?
Revolutions close with a total victory for one of the two opposing camps. Will that group ever say that the result of its victory has been something less than progress? That would be rather like admitting that they had been wrong and their opponents right. To them, at least, the outcome of revolution must be progress, and they are in an excellent position to make certain that future members of their community will see past history in the same way.6
To adapt the old saying: the victors write the science books.

Redesigned Goals

Gauch's name for this 'woe' does not fit well with how he describes it. Reassigned Role would be better.

This final woe is the natural result of all of the above. If science can't discover truth, or eliminate false theories, and only reflects the current preoccupations of a cloistered social group, then the high role it plays in modern thought is unjustified. Science is an ideological tyrant oppressing other ideas, and it's high time for us to drag science down from its stolen throne.

Paul Feyerabend's essay 'How to Defend Society Against Science' is very much worth reading, especially knowing the historical background I've given above. These same arguments pop up constantly in popular discussions when someone wants to shield an idea from scientific criticism (see Tim Minchin's poem Storm). I'll close out this post with Feyerabend's view on how government should treat science and remind you that this is a perfectly reasonable response if the 'woes' in this post are accepted:
Science is just one of the many ideologies that propel society and it should be treated as such (this statement applies even to the most progressive and most dialectical sections of science). What consequences can we draw from this result?

The most important consequence is that there must be a formal separation between state and science just as there is now a formal separation between state and church. Science may influence society but only to the extent to which any political or other pressure group is permitted to influence society.7

1. An example of falsification's entrenched role in popular debate:
2. Popper, K. (1935/2005). The logic of scientific discovery (3rd edition). Routledge. p. 317. (italics original; bolding added)
3. Salomon, D. (2006). Curves and surfaces for computer graphics. Springer. p. 141.
4. Hanson's own introduction of this idea is very readable:
5. Gauch, H. G., Jr. (2006). Scientific method in practice. Cambridge: Cambridge University Press. p. 84
6. Kuhn. T. S. (1962/1996). The structure of scientific revolutions. Chicago; The University of Chicago Press. p. 166.

No comments:

Post a Comment