Monday, March 23, 2009

Dude, where's my data?

At the risk of making it seem like I'm ranting at a dead horse, here is some more on the scientific status of social science, this time from JS Mill (1843, A System of Logic, Book VI, Chapter III).

His intro to the topic:

"1. [There may be sciences which are not exact sciences] It is a common notion, or at least it is implied in many common modes of speech, that the thoughts, feelings, and actions of sentient beings are not a subject of science, in the same strict sense in which this is true of the objects of outward nature. This notion seems to involve some confusion of ideas, which it is necessary to begin by clearing up."

True in 1843, true today.  Then an interesting analogy to the analysis of another complex system:

"Any facts are fitted, in themselves, to be a subject of science, which follow one another according to constant laws; although those laws may not have been discovered, nor even be discoverable by our existing resources. Take, for instance, the most familiar class of meteorological phenomena, those of rain and sunshine. Scientific inquiry has not yet succeeded in ascertaining the order of antecedence and consequence among these phenomena, so as to be able, at least in our regions of the earth, to predict them with certainty, or even with any high degree of probability. Yet no one doubts that the phenomena depend on laws, and that these must be derivative laws resulting from known ultimate laws, those of heat, electricity, vaporization, and elastic fluids. Nor can it be doubted that if we were acquainted with all the antecedent circumstances, we could, even from those more general laws, predict (saving difficulties of calculation) the state of the weather at any future time. Meteorology, therefore, not only has in itself every natural requisite for being, but actually is, a science; though, from the difficulty of observing the facts on which the phenomena depend (a difficulty inherent in the peculiar nature of those phenomena) the science is extremely imperfect; and were it perfect, might probably be of little avail in practice, since the data requisite for applying its principles to particular instances would rarely be procurable."

To conclude:
"Hence, even if our science of human nature were theoretically perfect, that is, if we could calculate any character as we can calculate the orbit of any planet, from given data; still, as the data are never all given, nor ever precisely alike in different cases, we could neither make positive predictions, nor lay down universal propositions."

This claim speaks to my earlier question of 'what is the point?'.  I do not doubt (in line with my commitment to the principles of physicalism and the causal completeness of physics) that there are fundamental laws of human behavior.  But I have no idea at what level these laws obtain.  It is dubious that these laws obtain at the level of aggregate collective phenomena like `democracies don't go to war with each other', especially if we are talking about any impressive level of accuracy.  And this means data collection problems, and so, Mill says, screw it.