From Supercrunchers to Disaster Algorithms

Decision making by numbers is gaining respect for accuracy and overall utility as Ian Ayres argues convincingly in Supercrunchers. In spite of the harsh commentary by a number of reviewers on Amazon, I insist that the subject of the book demonstrates the power and limitations of techniques for quantifying phenomena and using those results for public policy or business strategy. My rating of the book would award a 4 star for demonstrating the diverse ways in which regression analysis and randomized testing could be applied meaningfully.

IBM has constructed a model that is intended to guide in the management of disasters and their mitigation. I am unable to make much of its real capability from the description here but still consider that if it is deployed primarily in assisting with decisions on the responses that work in managing the crisis, then it is likely to be a very useful model.

However, it is also described as an instrument for automating complex risk decisions and begins to sound as if it is a glorified model that financial houses use in predicting what next year's bond prices will be. My reluctance is driven by the fact that the complexity of the mathematics used in the model is not necessarily correlated to the overall accuracy of the model. To what extent is th e construction of this algorithm different in approach to those that many hedge funds used to diminish risk. As an earlier post on hedge funds reported, John Kay thinks that a model is just as weak as the assumptions built into it in addition to the inherent weakness of the model itself.