I’ve recently finished reading Nassim Nicholas Taleb’s book The Black Swan and am slowly digesting the implications for the work that I do. “Black swan events” are occurrences that are unexpected (by the observer) but which have a disproportionately large impact on the observer. Taleb points to 9/11 as a classic example of a Black Swan event, insofar as it came out of the blue and led to a seismic shift in foreign policy thinking in the US and beyond, the reverberations of which are still with us. He also adds one other characteristic of Black Swan events: that once the event has occurred, it is rationalised by hindsight, as if it could have been expected. And, again, the media in the weeks following 9/11 was full of just such “I told you so” articles.
The problem, Taleb argues, is rooted in Western epistemology – the philosophy of the nature and scope of knowledge. The emphasis in Western thought since the age of the Enlightenment at least has been to draw general conclusions from specific observations. Modern scientists make heavy use of statistics to provide the links between observations and conclusions yet the statistics we use – and the normal distribution in particular – are good at predicting “average” or “typical” conditions, but are far less reliable at predicting consequences of low probability events – those that occur in the “tail” of the bell-shaped normal distribution curve. Yet the crux of Taleb’s argument is that it is often these events that have the most profound impact.
The main focus of Taleb’s ire are the financial analysts whose sophisticated models consistently fail to predict the various market-shaking events of the last thirty years or so. However, I saw much that was of at least tangential relevance to ecology. My fellow practitioners won’t like me saying this but ecology is a discipline which hovers between the “hard” and “soft” sciences. We can isolate particular components of the highly complex systems that we study and subject them to controlled experiments, following all the tropes of rigorous, quantitative science. Yet the very act of isolating components often means that we have created simplified, artificial situations and there is no guarantee that the components will act in the same way in the complicated and often unpredictable ecosystem from which they were plucked. So we often resort to soft science methods such as taking measurements in the field and then looking for correlations and associations. Our science advances from piecing together these various strands of evidence to make a convincing narrative.
Reading “The Black Swan” brought to mind a conversation I had had with colleagues a few years back. We were contemplating the ecology of submerged mosses in rivers. The data we had consisted of associations between these mosses and water chemistry. We were particularly interested in the consequences of reducing nutrient concentrations. Would this, we were wondering, lead to significant changes in the ecology of the river? As mosses often form a major part of the plant biomass in the rivers in which we were most interested, a change in these might have been expected. Some mosses, indeed, did inhabit rivers with higher nutrient concentrations than others. The question was exactly that which Taleb was posing for financial markets: did our models allow us to make predictions? Or, as they were based on correlations and associations with current conditions, were they descriptive, with very limited predictive power? The point that dawned on us was that one factor not included in our equations was the rare but catastrophic flood that would have the power to turn over the stable boulders which were the preferred substratum of the mosses and scour away many of the plants. These are the Black Swan events in rivers that would provide the tabulae rasa on which new species could establish themselves.
But this, in turn, runs into another set of problems: the need for “evidence-led policy”. Models which do not include the possibility of “Black Swan events” may have low predictive power and, therefore, an excuse for those policy makers and industry sceptics who want to avoid expenditure, especially where there is low confidence in a positive outcome. Yet, as Taleb, points out, the very nature of low probability, high impact events, means that it is difficult to build them into models. The dilemma is exacerbated by the short-termism which permeates many areas of science and science-policy.
So what is the answer? It is probably not bigger and better models, particularly if we are dealing with low probability events. The simplest solution may be to stop making inferences about ecological dynamics from spatial studies and shift the focus, instead, to monitor those situations where changes should be expected. There are numerous cases all over the country where water quality improvements are being enacted yet the ecological monitoring both before and after is often extremely limited. Yet it is only by following changes at these sites over a number of years – long enough to allow Black Swan events to exert their effects and with adequate spatial and temporal replication – that we will build up reliable evidence on which policy should be based. Long-term environmental monitoring by UK’s statutory agencies? Forget Black Swans: that would be a Pig’s Might Fly event.