Thursday, August 08, 2013

Fuelling physics envy?

Here’s an opinion piece I have just published in Physics World, before it was edited.


Physics envy might get a fresh stimulus from a new paper which claims to present “bibliometric evidence for a hierarchy of the sciences”. By analysing features such as authorship, mode of expression and range of citations in about 29,000 papers from maths to the social sciences and humanities, bibliometrics experts Daniele Fanelli of the University of Edinburgh and Wolfgang Glänzel of the Catholic University of Leuven say that there are good objective reasons to support the hierarchy that proclaims maths and physics the ‘hardest’ and most solidly grounded of the sciences [1].

The work provides a fascinating glimpse at the stylistic and methodological differences that exist between disciplines. It’s telling us something worth knowing: that fundamental differences in style and content across the sciences are real, so that it might be a mistake to evaluate and manage all the sciences in the same way. What it is not telling us is that physics is the most exemplary or exalted of the sciences.

The authors are mostly scrupulous in avoiding that implication. They suggest that this hierarchy is only to be expected because, progressing from physics to sociology, the complexities of the subject matter – the degrees of freedom, if you like – are increasing. So it is scarcely surprising that the phenomena become harder to interpret and consensus becomes harder to achieve: as Fanelli and Glänzel put it, the data “become less able to speak for themselves.” In this much, they are endorsing the view first espoused in the 1830s by the French philosopher Auguste Comte, who also posited a hierarchy from mathematics to physics, chemistry, biology, psychology and sociology based on the level of complexity involved. Comte was the father of positivism, which asserts that all authoritative knowledge derives from an objective, data-driven, scientific study of the world.

Comte’s hierarchy is typically expressed in terms of the ‘hard’ and ‘soft’ sciences. Fanelli and Glänzel embrace these terms, saying that they “seem to capture an essential feature of science”, and that pretending they do not exist could be a “costly mistake”. The authors don’t deny that all disciplines have cultural and “non-cognitive” components, but say that they seem nevertheless shaped “by objective constraints imposed by the subject matter”.

Before grappling with those assertions, let’s look at what the duo did. They figured that a defining characteristic of a ‘hard’ science is the ability to reach a shared interpretation of phenomena. Consensus might be expected to be reflected in several general features of papers. For example, they will be shorter, since there is less need to justify and explain a study; the references will tend to be more recent (key questions are resolved faster), fewer, less diverse and dominated by tightly focused papers rather than general monographs. But titles might be longer, since the issues addressed will be more precisely defined, and the number of coauthors might be greater, since more researchers share commonly agreed goals and because increased specialization makes collaboration essential. Fanelli and Glänzel analysed these parameters in thousands of papers on the Thomson Reuters’ Web of Science, categorized into disciplines such as physics, chemistry, plant and animal sciences, and psychiatry/psychology, and find that the expected trends are borne out by the data.

So what’s the problem? Let’s start with semantics: ‘hard’ and ‘soft’ are prejudicial terms. It is very difficult to avoid reading them both as “hard-headed/soft-headed”, suggesting that the social sciences are pervaded by woolly thinking, and as “hard/easy”, suggesting that the physical sciences are more intellectually challenging and reinforcing the snooty conviction that the most brilliant scientists choose physics. But arguably (most) questions in physics are in fact the easiest to answer securely because they tend to be the easiest to isolate and interrogate experimentally. Economics is failing to answer our real-world questions not because economists are less able, but because economics is so complex, with few if any universal laws and very patchy data. (There’s another reason too, which I’ll come to shortly.)

Yet more invidious than the ‘hard/soft’ terminology is the whole notion of a hierarchy. By definition, this implies a judgement of status: there’s a top and a bottom. At best it invokes condescension towards those disciplines unlucky enough not to be physics; at worst, we’re invited to feel impatient that these ‘softer’ sciences haven’t yet got themselves physics-ified. Comte certainly felt that all sciences aspire to the condition of physics, and he looked forward to the time when the social sciences reached this stage of higher evolution. It was in Comte’s time that historians of science began to construct the narrative in which the mathematization of nature, as displayed in Newton’s Principia, was the defining achievement of the Scientific Revolution, ignoring the fact that this approach was of no value in, say, zoology, botany, chemistry, geology and medicine. When Immanuel Kant declared that the chemistry of his day was “not science” because it was insufficiently mathematical, he was exposing his limited understanding of what chemistry was about, both then and now.

Not only is mathematization, with its consequent opportunities for reductive subdivision of problems, of limited value in some sciences, but they – the life and social sciences particularly – have a dependence on context and history that offers scant purchase for physics-style universal rules, and means different data sets may tell different stories. When those dependencies are neglected for the sake of simplification, as in mainstream neoclassical economic theory, the result is a model so abstracted and simplistic that no amount of empirical input – not even the near-collapse of the global economy – can make much impression on the ramparts of its ivory towers.

I happen to believe that many sciences, from biology to sociology, can in fact benefit from physics-based ideas. But placing physics at the top of the tree doesn’t help, because it blurs the view of where “physics thinking” is and isn’t appropriate. And presenting science in terms of “consensus deficit” is not just misguided but potentially dangerous. A quest for consensus tacitly accepts Comte’s assumption that all questions can be given a single, scientifically based answer. But many cannot, not just in the humanities but also in history, politics, ethics, the social sciences, economics and beyond. Even in the so-called ‘hard’ sciences, the value of having complementary but not entirely compatible models is under-rated. For some questions about humanity, we may be better served by a diversity of views – including old ones – than by a doomed dream of consensus.

1. D. Fanelli & W. Glänzel, PLoS ONE 8, e66938 (2013).

No comments: