HIGHLY OPTIMIZED TOLERANCE. Many natural and man-made systems exhibit power-law statistics. That is, when you plot the likelihood of an event (e.g., sizes of forest fires, power outages, and web file transfers, or losses due to hurricanes, floods, earthquakes, and man-made disasters) as a function of size the resulting graph will fall off proportionally to the size of the event raised to some exponent. Interactions or phenomena at many size scales (from very small to very large) contribute to the overall state of these systems. One theory which tries to explain all this is "self organized criticality." Jean Carlson of UC Santa Barbara (carlson@physics.ucsb.edu) and John Doyle of Caltech (doyle@cds.caltech.edu) now propose another theory, called highly optimized tolerance (HOT), which they believe does a better job of accounting for the tendency in interconnected systems to gain a measure of robustness against uncertainties in one area by becoming more sensitive elsewhere. As with energy conservation or the inexorable increase in entropy, efforts to violate the robustness principle will fail. Especially in biological evolution or in engineering, this means that a system might obtain robustness against common and designed-for uncertainties and yet be hypersensitive to design flaws or rare events.
For example, organisms and ecosystems exhibit remarkable robustness to large variations in temperature, moisture, nutrients, and predation, but can be catastrophically sensitive to tiny perturbations, such as a genetic mutation, an exotic species, or a novel virus. Engineers deliberately design systems to be robust to common uncertainties. Cost and performance tradeoffs force an acceptance of some hypersensitivity to (one hopes) rare perturbations.
In evolved or designed systems, this tradeoff leads to the "robust, yet fragile" characteristic of complexity, one feature of which is power laws. Doyle and Carlson have been exploring the application of their theory to a number of biological and engineering problems with the help of experts in those fields. (Carlson, Doyle, Physical Review Letters, 13 March 2000; Select Article; a longer version appears in Physical Review E, August 1999.)