Replies

"He said that if the system—which theoretically would analyze relationships among transactions such as credit card or airline ticket purchases—were applied to the entire population, almost as many people would incorrectly be identified as terror plotters as would be correctly fingered.

If it were that good, it would be fantastic. The innocent could easily be eliminated by a few discrete inquiries.

But the likely results are thousands of false positives, without a single correct hit, causing the FBI to waste hundreds of thousands of hours investigating innocent people.

I speak here as a programmer, not a civil libertarian.

I totally agree with both of you here.

There are plenty of systems able to flag unusual purchases and attempt to minimize theft. I have been phoned a few times during the past decade by my bank asking if I had made certain purchases they found somewhat unexpected for some reason. I was happy to tell them "yes" and happy to see attention was paid to such things. It reduces my credit card cost.

Any system able to determine "terrorist" purchases with a reliability even close to 1:1 true:false rate would be extremely good to develop. Personally, I think it would be criminal for us to neglect using any system we could get better than 1:10. Your points about individual privacy being protected by the overwhelming numbers are well spoken.