Monday, January 11, 2010
The Black Swan
On the merits, I disagree with Taleb that the Black Swan poses anything like an archetypal problem for epistemology, prediction or forecasting. All it seems to me to require is constant updating (and sometimes replacement) of theories in light of new information. The empirical assertion "All swans are white" was, in Popper's sense, falsified by the first black swan to be observed. But that observation hardly proved that making empirical predictions is a fruitless enterprise. For example, we can still predict quite accurately that most swans are white. In fact, we can know make fairly precise predictions of the proportion of swans that are white based on our knowledge of genetics and probability theory. In short, prediction improves with the quantity and quality of data. Forecasting errors are inevitable but reducible over time.
This leaves untouched Taleb's larger claim that everything interesting or important that happens deviates from what is known or predicted, but that is really just a truism. We are always more likely to be impressed (and upset) by unexpected storms than those the weather forecasters accurately predicted. There's no denying, however, that the advent of computers and radars has made short-term weather forecasts generally more accurate.
Posted by Daniel H. Cole at 10:30 AM