You’re not a molecule, but sometimes you’re a statistic
The editorial in the latest issue of Nature, written by me, could in its edited form (my original draft is below) seem to present a capitulation to the view of social science advocated by Steve Fuller at Warwick, who has previously been highly critical of the statistical perspective discussed in my book Critical Mass. (My response to Fuller is here.) But that’s not really how it is. The pull quote (“The goal of social science is not simply to understand how people behave in large groups but to understand what motivates individuals to behave the way they do.”) is equally true in reverse, which is what Fuller seems blind to. I’m more that happy to make explicit what statistical ‘laws’ overlook. But to deny that group behaviour matters, or that it can differ from that predicted by linear extrapolation from individuals, is to deny the ‘social’ in social science, which seems to me a far more egregious oversight.
Fair point from the editor, though: it wouldn’t actually be hard at all to improve on Mill’s words, in the sense of leavening the Victorian stodge. But I hope the editorial doesn’t now seem to be implicitly critical of the González et al. paper that motivated it, on the grounds that it focuses on the masses and not the individual. This paper does reveal information about both. That very issue, however, has provoked an absurd level of hysteria in the wake of the news story we ran. It seems some people who haven’t bothered to read the paper are concerned about privacy. Makes you wonder what they have to hide (not that anyone would be finding out in any case, given that the data were rendered anonymous). Do these people ever stop to think what is happening to the data every time they make a purchase on their credit cards?
“Events which in their own nature appear most capricious and uncertain and which in any individual case no attainable degree of knowledge would enable us to foresee, occur, when considerable numbers are taken into account, with a degree of regularity approaching to mathematical.” It would be hard to improve on John Stuart Mill’s words to encapsulate the regularities found in human mobility patterns on page 779 of this issue. Who would have thought that something as seemingly capricious as the matter of where we go during our daily lives could yield such lawfulness?
One of the remarkable features of this work is not the results, however, but the methodology. Social scientists have long struggled with a paucity of hard data about human activities – social networks, say, movement patterns. Self-reporting is notoriously unreliable and labour-intensive. The use, in this case, of mobile phone networks to track individuals has supplied a data set of proportions almost unheard of for such a complex aspect of behaviour: over 16 million ‘hops’ for 100,000 people. The resulting statistics show a strikingly small scatter, giving grounds for confidence in the mathematical laws they disclose.
This adds to the examples of information technologies offering tools to the social scientist that provide a degree of quantification and precision comparable to the so-called ‘hard’ sciences. Community network structures can be derived from, say, email transmissions or automated database searches of scientific collaboration. Online schemes can even enable genuinely experimental study of behaviour in large populations, complete with control groups and tunable parameters.
Making sense of these data sets may require a rather different set of skills from the conventional statistical approaches used in the social sciences, which is why it is no surprise that studies like the present one are often conducted by those trained in the physical sciences, where there is a long tradition of investigating ‘complex systems’ of interacting entities. One view might be that this lends some prescience to the suggestion of sociologist George Lundberg in 1939: “It may be that the next great developments in the social sciences will come not from professed social scientists but from people trained in other fields.” Lundberg was a positivist eager for his field to adopt the methods of the natural sciences.
The ‘physicalization’ of the social sciences needs to be regarded with some caution, however. While some social scientists aim to understand the ways people behave in large groups, others insist that ultimately the goal is not to uncover bare statistical laws and regularities but to gain insight into what motivates individuals to behave the way they do. It is not clear that universal scaling functions can offer that: however vast the data set, the inverse problem of deriving the factors that produce it remains as challenging as ever. Statistical regularities may conjure up images of Adolphe Quetelet’s homme moyen, the ‘average man’ who not only tends to deny the richness of human behaviour but even threatens to impose a stifling behavioural norm.
It would be wrong to imply that the interest of these findings is restricted to the conventional boundaries of the social sciences. Epidemiologists, for instance, have traditionally been forced to work with very simple descriptions of dispersal and contact, for example based on diffusive models, for lack of any hard evidence to the contrary. But recent work has made it very clear that the topology and quantitative details of contact networks can have a qualitative impact on the transmission of disease. There is sure also to be commercial interest in information about patterns of usage for portable electronics, while the nature of mass human movement could inform urban planning and the development of transportation networks.
But for the social sciences proper, the latest results suggest both an opportunity and a challenging question: how much of social behaviour do we capture in statistical regularities, and how much do we overlook?