A few years ago, there seemed to be a risk that the Large Hadron Collider (LHC) at CERN would destroy Earth. The risk was based on speculative, if serious, physics. The probability that Earth would be destroyed seemed low because the theories that allowed trouble were speculative. Nevertheless, even given a small probability, destruction of Earth would have an enormous negative expected value (probability times value, the appropriate metric for decision theory).
Collider advocates put forth several reasons not to worry, but others found lacuna in those reasons. CERN conducted two safety studies, the second necessitated by flaws in the first. The final study was fairly good, but was still subject to criticism. Despite all this, CERN declared that the probability that the Earth would be destroyed was zero and fired up the LHC in 2008. Despite a minor industrial accident, the Earth was not destroyed.
Much of the risk that we can control is behind us. In most models destruction of Earth takes a few years, but there is nothing we can do if the process has been initiated. Continued collider operation poses residual risk from events that happen infrequently and have not happened yet. There is also risk from CERNís planned upgrade, and from the next generation of colliders.
Against this risk, we have to balance the probability that LHC research will discover something that will save us from other risks, something that would not be discovered by other methods.
Discussion of the problem
Forum (Letters and articles published here)
References and links to published articles
Book writing project: Management of Positive and Negative Singularities: A Socratic Dialog