This morning, Ben Goldacre, the Guardian columnist and Bad Science blogger) has written on the difficulties and dangers of datamining to identify criminals and terrorists (Datamining would be lovely if it worked).
He's written this on the back of an IPPR report (The National Security Strategy: Implications for the UK intelligence community), and he's raising issues of accuracy and false positive rates - clearly analogous to evidence-based medical trials, which are closer to his usual blog fare. Goldacre quotes the author of the report: