All I want for Christmas is the presumption of innocence
by Matthew Gast
Last week, Andy Oram
wrote about a recent INS directive for men from twenty Muslim or Arab countries to report for screening. The
New York Times article he links notes that "all but 20 of the hundreds" of detainees were released. Defining "potential terrorist" as "male of Arab descent or Muslim heritage" is so antithetical to civil liberties that a round-up is unacceptable. Even if we were willing to throw our civil liberties out the window (which, for the record, I am not), mass screening is a costly and ineffective approach.
The official used as the source for the Times story declined to state how many people were processed. Naturally, that made me curious. Although I am not a serious statistician, "all but 20" has much less relevance without knowing what it is measured in relation to. Did the INS harass thousands of people, most of whom were undoubtedly honest, to find the 20 suspects worthy of further investigation? Probably, given the broad language being used to describe the effort elsewhere. (I first heard about it on KQED's
California Report on December 18. There was a follow-up story December 23 reporting on related protests.)
While pondering the difficulties in rounding up such a broad group of people, and wondering whether or not they were treated as humanely as the INS insisted, I came across the linked message on the Interesting People mailing list. The author, a computer science professor at UT-Austin, has an incisive analysis of the statistical failures of a broad screening approach, even if supervised by humans. His argument is constructed by analogy to screening for a rare disease in a medical laboratory. No screening process will identify everybody who has the condition, and substantial numbers of unaffected people are misidentified along the way.
Moreover, in many lab settings, you can correctly assume that you have all the samples. In other cases, extremely strong privacy protections can help people feel less wary about the privacy implications of tests. In the case of an INS order to report for a background check, it is safe to assume that people with questionable backgrounds will attempt to avoid reporting, thus skewing the population of those screened far more in favor of the innocent. Such a tilt will tend to make the cost/benefit calculation worse by increasing the costs to innocent people while decreasing the small benefit that results from hysterical mass screening.
The presumpution of innocence has long served as a check on intrusive governments, and it is one of the pillars of our criminal justice system. Demagogues often attempt to weaken it by presenting a false dichotomy between the abstract concept of rights and a tangible feeling of security, often conjuring up criminal demons to frighten us into choosing the latter. Screening programs targeting broadly shared characteristics must be resisted on the grounds that they are antithetical to this country's founding principles. Furthermore, as a practical matter, they are almost totally ineffective at anything other than wasting money while proving our hypocrisy.
Speaking of hysterical mass screening and wasting money, the TSA apparently has decided to follow the lead of whatever bonehead at the INS ordered the round up. The TSA
advising fliers to leave luggage unlocked to make supplmental baggage screening easier. The same mathematical analysis in searching for rare conditions almost certainly applies, given that most airline passengers are not terrorists. I would guess that most checked bags are not selected for a supplemental hand-search, and of those that are, most are not problematic. Meanwhile, there are definite costs associated with leaving checked luggage unlocked, especially at a time of year when people are likely to be transporting valuables. Fortunately, the TSA is looking out for me and is willing to trade the security of my possessions for my safety.
Fine, I agree. So -- what _should_ we do???
I'm not much of a statistician, either, but I
understand the Bayesian statistics enough to
agree with you. Broad screening for a rare condition
will lead to a large number of false positives, while
providing no guarantee that all the rare conditions
are caught. (Unless, of course, the test is perfect :-)
Why do anything?
It is the perception of bigotry and self-righteousness which has caused so much distrust and hatred toward the US and the UK (my country) throughout the World. This kind of behaviour will cause even more anger and sadly also achieve nothing.
Yet more cause for alarm
Thanks, Matthew, for refering back to my weblog. Here's another bold expose, this one from the Washington Post: