Letter: Terrorism database problems don't add up


Regarding "Lawmaker: Terrorism info database troubled": I've been involved in statistics most of my life. I'm 64.

There are mathematical problems in the terrorism database that need to be considered along with the possible administrative ones. Fixing the wrong problem won't help, and fixing one problem when there are more than one problem affecting the result may not help much.

The main mathematical problem I am talking about is the problem of low relative frequency. It makes statistical prediction extremely difficult.

Suppose 200 million people fly in the United States in a year. Suppose further there are 1,000 suicidal terrorists in the country. Suppose you would be satisfied to catch 3/4 of them and you achieve a remarkable 90 percent accurate prediction system. You would still have 250 suicidal terrorists walking past the screeners with only the same inspection that most of the 200 million folks get. Perhaps worse, you would then have 19,999,750 persons "identified" by the system as apparent terrorists who really are innocent like the rest of us.

What would you do with those false positives? I'm not sure you could accomplish much in an hour, but if you spent one hour each on them, you would be spending more than 10,000 staff years a year on them. At $40,000 a year each, that's $400 million a year to establish the harmlessness of the harmless folks the system erroneously identified as terrorists.

Worse than all that, there is no guarantee that the system would work even that well. As soon as you push the "start" button, the real terrorists will start work to defeat the system. How hard will it be for them to figure out how to fake an identity, use a constant stream of new recruits, work through innocent dupes, make the guilty look innocent, and so on?

It's easy to point fingers at management problems. Every organization has them; they are human problems. It is inherently difficult to work the kinks out of a system designed to deal with rare events, and these are the rarest of events.

It is likely that the problems being "identified" are not just management problems, but problems that result from errors inherent in the design assumptions. I do not see that they can be solved. So far the approach seems to be to ignore the thing we cannot address. That won't help. For the system to work, all the elements have to work.

Ronald Hietala

What do you think? Paste a comment in the box below (registration required), or send your comment to letters@fcw.com (subject line: Blog comment) and we'll post it.

Featured

Reader comments

Please post your comments here. Comments are moderated, so they may not appear immediately after submitting. We will not post comments that we consider abusive or off-topic.

Please type the letters/numbers you see above