Taleb and Triviality: A Sociological View

October 14th, 2008

Views: 563

Posted by ChrisG at 3:03 pm

Chris Dillow offers a re-interpretation of Nassim Nicholas Taleb’s treatment of the problem of induction, which increasingly seems to have been totally absorbed by the “Black Swan” brand, suggesting along the way that there is nothing particularly interesting about Taleb’s theory that “wild uncertainty” is the norm, and a predictable future is an unusual phenomenon indeed.

My own reaction to Taleb’s book was not, I suspect, unusual – his epic verbosity and tedious arrogance have so far conspired to defeat three attempts to read it all the way through.

And as with so many popularizing macro-trend books, the swarm of anecdotes the author deploys tend to bob and weave their way towards what are ultimately trivial points. But in addition, for a sociologist interested in risk and uncertainty, it’s clear that Taleb is not saying anything that various writers in the academic sociology of scientific knowledge (SSK) have not been saying for at least a couple of decades – and with, for that matter, far more insight to offer into the social processes and conditions which make blindness to uncertainty so prevalent. For his part, Taleb persists in the methodological individualist’s (or economist’s) habit of putting it all down to psychological biases (the “narrative fallacy”, etc. etc.).

One sociological reason for the prevalence of uncertainty blindness is the non-Bayesian view, persistently common within some institutions that make heavy use of risk analysis, that risk is somehow an objective phenomenon. This harks back to the economist Frank Knight’s interpretation of the difference between risk and uncertainty back in the 1920s, where uncertainty appeared to be whatever was left over after the calculations of risk based on past observations of a given system had been produced. This meant, for example, that the probability of an event dependent on an interaction between well-observed system A and as-yet-unmapped external system B was not a well-defined risk but an ill-defined uncertainty. The uncertainty of such an event, however, derived solely from our current lack of knowledge about B. For this view of risk, uncertainty can in principle be progressively tamed – it’s the Whig view of uncertainty, if you like.

However, what some sociologists of science offer is a less sanguine view of uncertainty – basing it on fundamental, or if you like, existential ignorance, which in turn derives from the inherent perspectivism of empirical research. Empirical research is always theory-laden, sure; but what this means is that the selection of what to study provides the conditions under which (in the typical idiom of SSK) knowledge is produced. But alongside this knowledge, ignorance is also produced. It’s not like ignorance (and uncertainty) are a remainder waiting to be removed, in time, by assiduous research. Rather, the pursuit of certain avenues of research (which reflect, in turn, pre-existing social priorities) treats certain other potential avenues as uninteresting, barren or impossible to investigate with current methods, instruments etc.

In this way, the frame within which research explores the world is highly conditioned, and is not fully under the control of those who want to do the research. It is given its characteristic shape by external social dynamics. Uncertainty, on this view, is therefore an objective effect of interactions between humans and their world, out of which entirely unforeseen surprises can originate from even relatively banal interactions of familiar and well-understood processes, such as in the case recounted by Ian Hacking of interference effects arising from the use of baffles inside cooling-tower chimneys for preventing the emission of ash.

An example from one of the foremost representatives of this kind of sociological approach to uncertainty, Brian Wynne, illustrates how ignorance and uncertainty are manufactured. Following Chernobyl, the persistence of radiocaesium in sheep living on the Cumbrian hills in the UK was found to contradict earlier scientific predictions about its environmental persistence. The reason for this, Wynne argues, was that the measurements (dating from the 1960s) on which the earlier predictions were based had been guided by a particular set of social assumptions about the kind of risk-scenario that would be relevant in the case of environmental contamination by radiocaesium (i.e. a dose of radiation to an individual standing on the ground in a contaminated area). But after Chernobyl, a different exposure scenario (contamination via the food chain from grazing sheep to humans who ate them) became the focus of attention, and different parameters of the system under observation (the chemical mobility of radiocaesium as well as its physical disposition in the soil) were measured as a result. In this case, the re-focusing of policy interest on different modes of exposure served to render previous scientific research on risk obsolete. Changing social priorities showed up what had been unacknowledged limitations on scientific knowledge.

The point then is perhaps that uncertainty is produced unpredictably by the processes through which empirical knowledge is arrived at, but that the legitimacy of social institutions often rests on well-established processes by which the objective nature of this uncertainty is denied (both within research and within contexts in which that research is used and interpreted). There is more than psychology at issue here: uncertainty has an institutional politics, in which demonstrable (or at least temporarily unfalsifiable) claims of control provide a source of authority.