Math...harder than it looks.

Kevin Drum indulges in a bit of wishful thinking and hopes that next-generation lie detectors can help catch terroists in airports.

Even in a test environment, the machine missed 15% of "terrorists" and incorrectly flagged 8% of innocent travelers. Cogito's goal is to get that to 10% and 4%, which is still far too high to be of very much value.

This technology isn't ready for prime time yet, not least because sticking people inside an MRI machine at airports isn't exactly a feasible concept. But I wouldn't be surprised if a better and more reliable version of this technology were available within five to ten years. If, say, it could replicate Cogito's planned accuracy, and the two were combined, the resulting device might have a 99% chance of catching terrorists and only a .1% rate of false positives. That would be genuinely useful.

Let's see whether this holds up to a quick analysis. The 99% end looks pretty good. Even a large terror cell with say 2 dozen memebers would be unlikely to get a member through. Looks promising. However, a 0.1% rate of false positives is 1/1000. That sounds pretty good at first, but a decently sized airport like BWI gets 54,088 passengers a day. That's 54 potential terroists to process every day, from a single airport. So you're looking tens of thousands of suspects per year, just from one airport. Multiply that by a few dozen airports, and you're in for a pretty big haystack to search for that same 24 member terror cell. How many hundreds of thousands of people a year can we delay, detain, or investigate?