Abstract: (we have) “a sufficiently strong propensity not only to make divisions in knowledge where there are none in nature, and then to impose the divisions on nature, making the reality thus conformable to the idea, but to go further, and to convert the generalisations made from observation into positive entities, permitting for the further these artificial creations to tyrannise over the understanding.”
Henry Maudsley 1867 “The Physiology and Pathology of the Mind”
The purpose of this talk is to examine how popular conceptions and measurements of risk can lead us into serious errors of judgement. A particular illustration will be the “Value at Risk” measure used to such little effect in Basel 2 capital requirements of Banks, and lessons relevant to Solvency II regulation of insurance companies coming into force in October 2012.
The thrust of the talk is that we can assess risk using three principles - first, framing the problem to allow for unforeseen or excluded possibilities and their change in time; second, use of different models to estimate the range of error in the measurement of risk; third, setting a level of acceptable risk relative to these measures. These principles are of practical value in our daily lives not just esoteric matters for scientists or financial wizards. For instance, we are all familiar with activities such as smoking whose risks were perfectly acceptable in one era but considered beyond the pale in another.
Particular topics covered in the talk will include why “Knightian uncertainty” is a false distinction, why the oft-heard statement that “all models are wrong” misses the point, why time is the crucial dimension of real-world risk, and failures of risk management models in the recent financial crisis.
The title of the talk is an echo of Stephen Jay Gould’s book “The Mismeasure of Man” where he argued that the idea of a single number, the IQ, as a measure of man’s capability was a misguided simplification. The talk will show how the fallacies of reification and ranking identified by Gould apply also to the measures of risk and with similar consequences of incorrect judgement and action. As Gould did with IQ, the talk will also examine the importance of context in framing the problem and the actions taken as a result of the measured values.
The talk should be of interest to a general audience, with particular areas of special interest to scientists, economists, philosophers, and those in the financial services sector interested in risk.
Dr Peter Taylor is a Research Fellow concentrating on the area of risk with the Program on Ethics of the New Biosciences and James Martin School. Peter spent 25 years working in the Lloyd's insurance market where he has managed IT and loss modelling departments and led and participated in many projects. He has been a director of insurance broking and underwriting companies and market organisations, and helped to establish the Lighthill Risk Network, a non-profit organisation that brings together the business and scientific communities for their mutual benefit. Peter is still an active consultant in the City of London, but spends as much time as he can working with the Program. Peter has a long-standing interest in all aspects of risk, whether in insurance or in science generally, particularly the practical application of the theory of risk, and the analysis of emerging risks. Peter has a background in the foundations of quantum theory for which he was awarded his D. Phil at Oxford, and in July 2007 organised the Everett@50 Conference at the Philosophy Centre in Oxford