More on Philip Kitcher’s brilliant essay on scientism (“The Trouble with Scientism,” The New Republic; see my earlier comments here).
Kitcher writes:

The problem with scientism — which is of course not the same thing as science — is owed to a number of sources, and they deserve critical scrutiny. The enthusiasm for natural scientific imperialism rests on five observations. First, there is the sense that the humanities and social sciences are doomed to deliver a seemingly directionless sequence of theories and explanations, with no promise of additive progress. Second, there is the contrasting record of extraordinary success in some areas of natural science. Third, there is the explicit articulation of technique and method in the natural sciences, which fosters the conviction that natural scientists are able to acquire and combine evidence in particularly rigorous ways. Fourth, there is the perception that humanists and social scientists are only able to reason cogently when they confine themselves to conclusions of limited generality: insofar as they aim at significant — general — conclusions, their methods and their evidence are unrigorous. Finally, there is the commonplace perception that the humanities and social sciences have been dominated, for long periods of their histories, by spectacularly false theories, grand doctrines that enjoy enormous popularity until fashion changes, as their glaring shortcomings are disclosed.
These familiar observations have the unfortunate effect of transforming differences of degree into differences of kind, as enthusiasts for the alleged superiority of natural science readily succumb to stereotypes and over-generalizations, without regard for more subtle explanations. Let us consider the five foundations of this mistake in order.

On scientism’s first problem — “the sense that the humanities and social sciences are doomed to deliver a seemingly directionless sequence of theories and explanations, with no promise of additive progress” — Kitcher elaborates:

The most obvious explanation for the difficulties of the Geisteswissenschaften, the humanities and the study of history and society, is that they deal with highly complex systems. Concrete results are often achieved in particular instances: historians and anthropologists are able to be precise and accurate by sacrificing generality, by clear-headedly disavowing the attempt to provide any grand overarching theory. No large vision of history emerges from our clearer understanding of the bombing of Dresden, but the details are no less powerful and significant. In this respect, moreover, matters are no different in the natural sciences. As we shall see, science often forgoes generality to achieve a precise and accurate answer to an important question.
In English we speak about science in the singular, but both French and German wisely retain the plural. The enterprises that we lump together are remarkably various in their methods, and also in the extent of their successes. The achievements of molecular engineering or of measurements derived from quantum theory do not hold across all of biology, or chemistry, or even physics. Geophysicists struggle to arrive at precise predictions of the risks of earthquakes in particular localities and regions. The difficulties of intervention and prediction are even more vivid in the case of contemporary climate science: although it should be uncontroversial that the Earth’s mean temperature is increasing, and that the warming trend is caused by human activities, and that a lower bound for the rise in temperature by 2200 (even if immediate action is taken) is two degrees Celsius, and that the frequency of extreme weather events will continue to rise, climatology can still issue no accurate predictions about the full range of effects on the various regions of the world. Numerous factors influence the interaction of the modifications of climate with patterns of wind and weather, and this complicates enormously the prediction of which regions will suffer drought, which agricultural sites will be disrupted, what new patterns of disease transmission will emerge, and a lot of other potential consequences about which we might want advance knowledge. (The most successful sciences are those lucky enough to study systems that are relatively simple and orderly. James Clerk Maxwell rightly commented that Galileo would not have redirected the physics of motion if he had begun with turbulence rather than with free fall in a vacuum.)
The emphasis on generality inspires scientific imperialism, conjuring a vision of a completely unified future science, encapsulated in a “theory of everything.” Organisms are aggregates of cells, cells are dynamic molecular systems, the molecules are composed of atoms, which in their turn decompose into fermions and bosons (or maybe into quarks or even strings). From these facts it is tempting to infer that all phenomena–including human actions and interaction–can “in principle” be understood ultimately in the language of physics, although for the moment we might settle for biology or neuroscience. This is a great temptation. We should resist it. Even if a process is constituted by the movements of a large number of constituent parts, this does not mean that it can be adequately explained by tracing those motions.
A tale from the history of human biology brings out the point. John Arbuthnot, an eighteenth-century British physician, noted a fact that greatly surprised him. Studying the registry of births in London between 1629 and 1710, he found that all of the years he reviewed showed a preponderance of male births: in his terms, each year was a “male year.” If you were a mad devotee of mechanistic analysis, you might think of explaining this — “in principle” — by tracing the motions of individual cells, first sperm and eggs, then parts of growing embryos, and showing how the maleness of each year was produced. But there is a better explanation, one that shows the record to be no accident. Evolutionary theory predicts that for many, but not all, species, the equilibrium sex-ratio will be 1:1 at sexual maturity. If it deviates, natural selection will favor the underrepresented sex: if boys are less common, invest in sons and you are likely to have more grandchildren. This means that if one sex is more likely to die before reaching reproductive age, more of that sex will have to be produced to start with. Since human males are the weaker sex — that is, they are more likely to die between birth and puberty — reproduction is biased in their favor.
The idea of a “theory of everything” is an absurd fantasy. Successful sciences are collections of models of different types of phenomena within their domains. The lucky ones can generate models that meet three desiderata: they are general, they are precise, they are accurate. Lots of sciences,natural sciences, are not so fortunate. As the ecologist Richard Levins pointed out decades ago, in many areas of biology–and, he might have added, in parts of physics, chemistry, and earth and atmospheric science as well–the good news is that you can satisfy any two of these desiderata, but at the cost of sacrificing the third. Contemporary climatology often settles for generality and accuracy without precision; ecologists focusing on particular species provide precise and accurate models that prove hard to generalize; and of course if you abandon accuracy, precision and generality are no problem at all.

The tension between generality, accuracy, and precision is inherent to all human knowledge. Ideally, knowledge should be general (a concise insight should apply to a broad spectrum of reality), accurate (the insight should tell the truth about reality) and precise (the insight should be specific to the reality).
Differences in generality, accuracy, and precision are inherent to different fields of knowledge. As Kitcher notes, historians often sacrifice generality for accuracy and precision. Overarching simplifications of extraordinarily complex historical events often detract from accuracy and specificity, and are wisely eschewed. Novelists and playwrights eschew accuracy (their characters and events do not actually exist) in order to tell the broader truth (generality and precision) about human affairs.
The same tension exists in science. In some sciences, such as physics, generality and precision are important, although accuracy may need to be sacrificed (physicists have struggled for a century with the inadequacy of “particle” or “wave” to accurately describe subatomic reality). “Stamp collector” sciences such as paleontology and cladistics employ accuracy and precision, understanding that impertinent claims of generality may bias the record. Evolutionary biologists often run afoul of this basic principle in their “stamp collecting” science, by asserting broad meanings to the actual record that are not supported by the evidence. Witness the tumult over “punctuated equilibrium,” which is a generalized inference fabricated to evade the incongruity between the fossil record and the gradualism inherent to Darwinian theory, itself another fabricated generality.
Inappropriate inference to a kind of knowledge not inherent to reality is the fundamental error of scientism. Consider the study of the movements of the keys on my keyboard as I type this. The most basic study would be of the physics of the keys — how they respond to the pressure of my fingertips, what electrical currents and potentials are generated by the contacts they make, etc. This study — Newtonian mechanics, electrodynamics, computer science — can be quite accurate and precise, but it is hardly general. This is because the pattern of my key strikes is also conveying ideas via a language, and the study of ideas and languages is a matter of philosophy and linguistics, to which physics cannot generalize. To generalize the physics of the keystrokes to the philosophy conveyed by the keystrokes is not to gain deeper insight into my keystrokes. It is simply to make a stupid logical error. One cannot understand the ideas I am conveying by doing physics.
The application of generality, accuracy, and precision inherent to a field of knowledge is determined by the nature of the field of knowledge itself. As long as one respects the inherent constraints on knowledge in a particular field, no field is “better” than any other. The scientistic mistake is to run afoul of the constraints. Scientism is a misrepresentation of reality.
Michael Egnor also blogs at Egnorance.