Faulty risk analysis

Attempts to predict the chances of technological disasters have an interesting if sordid history. NASA rocket experts developed in the 1960s the pioneer version of risk assessment. Their mathematical models ('fault trees') predicted the failure rate of a particular rocket stage to be 10^-4 per demand i.e. one failure (on average) per 10,000 launches. After two failures in the first 100 tests, NASA abandoned the attempt to predict absolute failure rates of complex systems such as rocket engines, and relegated fault-tree analysis to the more modest role of comparisons between similar designs.

The involvement of NASA in the genesis and abandonment of 'absolute probabilities' through fault-trees was documented by the Union of Concerned Scientists some years later.

Its creators having thus abandoned it, the nuclear reactor enthusiasts then misused this discredited method to produce the 'one in a million' slogans of the AEC 'Rasmussen/Levine' report. Next, the LPG (liquified petroleum gas) industry took over these discredited 'one in a million' mottos, and this type of confusing pseudo-scientific PR remains a lucrative trade right down to the present day, notably in the hands of consultants such as Science Applications Inc. and Arthur D Little Corp. But, a decade ago, the leading Australasian experts in applied mathematics testified on oath in New Zealand that no useful accuracy was available by these obfuscatory formulae.

Attempts at risk analysis for GE are, obviously, doomed to be even more misleading. The system of a living cell, even if no viruses or foreign plasmids (let alone prions) are tossed in, is incomparably more complex than a nuclear reactor. There is no prospect of imagining most of the ways it can go badly wrong. Evidently, one cannot begin to estimate the risk of a mishap the qualitative form of which has not yet been imagined. We do not and cannot know how to put useful, justified numbers on the chances of severe GE mishaps.

You do not have to be a rocket or GE scientist to understand these general truths about biology. Many gene-splicings come to naught; some others may yield only the desired outcome; but the few major mishaps, as with nuclear power, dominate the assessment so as to rule out this approach to science and life.

For more about risk analysis, see: