The probability of not suspecting David Headley

A 250-fold improvement in obfuscation

In Superfreakonomics, the sequel to their first book, Steven Levitt and Stephen Dubner write that in the United Kingdom, Muslim names “would turn out to be one of the strongest demographic markers for the algorithm” to identify people who might be terrorists.

A person with neither a first or last Muslim name stood only a 1 in 500,000 chance of being suspected a terrorist. The likelihood for a person with a first or a last Muslim name was 1 in 30,000. For a person with a first and last Muslim names, however, the likelihood jumped to 1 in 2,000.

All this suggests that if a budding terrorist wants to cover his tracks, he should go down to the bank and change the name on his account to something very un-Muslim (Ian, perhaps). [Superfreakonomics]

Evidently, Daood Gilani, 49, a US citizen of Pakistani extraction, knew his freakonomics. His chances of being suspected went down from 1 in 2000 to 1 in 500,000, perhaps even lower given his US citizenship, when changed his name to David Coleman Headley.

6 thoughts on “The probability of not suspecting David Headley”

  1. Good one!

    Apparently terrorists didn’t need mathematical models to understand this. Why don’t more do it? Also, it is a way to beat border crossing profiling.

  2. Wow.. Dubner and Leavitt have thrown political correctness to the winds.. already they angered the environMENTALISTS by questioning their cap and trade nonsense over “global warming” or climate change or whatever the hell its called today..

    And now they have taken on the holy cow of Western Political Correctness post 9/11 – these guys need to watch their back.

  3. Two types of error can occur in any statistical pattern recognition, or to use the politically loaded term, in profiling.To illustrate, in the context of Islamic terrorism, the algorithm may reject “David Headley” as a potential terrorist, when he is actually a terrorist – false negative or Type 2 error, and may identify “Abdul Kalam” as a terrorist , when he is actually not – false positive or Type 1 error. Well-designed statistical algorithms aim to minimize both these errors. Often, these are sophisticated learning algorithms – they learn from prior errors.

    What the prediction of Levitt and Dubner implies is that Muslim first and last names will be given high weight in the algorithms to come. Why? Higher weights for these names would have resulted in fewer errors. That does not mean, however, that weights for other features play no role in the classification. They do. Take a look at how many clues about Maj. Nidal Hasan’s behavior and attitudes were missed or overlooked.

    It makes common sense, too. For every David Headley, there could be, and there are, hundreds of Nidal Hasans, Khalid Mohammad Sheikhs, and Ajmal Kasabs. It’s not easy to change names, at least not here in the United States. There are court records of prior names and all that. Background checks are quite thorough, if only the politically correct refrained from intervening. Today, if a Muhammad Ibrahim were to walk into a Bank to change his account name to Ian Smith, all sorts of alarm bells will be turned on, if it is at all feasible. Even the terrorists have only limited resources, and they are more likely to be expended on a few puppeteers, not the puppets. If I know this, I expect the algorithm designers to know a lot more.

    I don’t mean to lecture on statistics, but it is important to separate politics from science. Failure to do so will be costly in terms of lives lost, as in Fort Hood.

  4. @Rational Fool

    Thanks for the interesting insight provided, but what is the exact point you’re making? Are you supporting the freakonomics observation?

Comments are closed.