On complexity, rare events and the systemic underestimation of risk.

I'd thought it was inscrutable mathematical voodoo, so never bothered reading further, but it amounts to determining the value of a particular piece of evidence by taking into account both false positives and false negatives. Simple but powerful.