When the Profile Fits the Crime

[Kurush]

I see the profiling fans still aren't reading the Carnival Booth paper, which addresses their arguments. I'll summarize and grossly oversimplify:

Scenario 1:
There are 10 terrorists in a cell. They successively go through security checkpoints, with no suspicious items at all, to test whether they are selected for heightened checks. 5 of the terrorists fit a profile and are checked every time. Five others them do not fit a profile and are not checked. The first 5 terrorists prepare the bombs and then leave the target nation.

The remaining five enter the target area. Because they don't fit a profile, they have a 0% chance of being thoroughly checked. They successfully set off the bombs.

Scenario 2:
Again, 10 terrorists. They probe the checkpoints but find no pattern. They prepare the bombs, and try to pass the checkpoints to set them off. If every third person is randomly checked, there is a 98% chance that at least one of the bombers will be caught. The target zone is shut down and secured by police. Some of them set off their bombs at the checkpoints, others are caught, but assuming reasonably well designed checkpoints, little loss of life will occur.

In other words, profiling never works against terrorist cells who test security beforehand. It is a well established fact that Al-Qaeda's MO is to extensively test security, so profiling systems are completely useless.

 

[cuchulainn]

Iran is an Islamic theocracy, last time I checked.

Yes it is, and Iranians are Caucasians. That's why we should include Caucasians in our profiling. Iranians look like southern Europeans (Italy, Spain, etc). Iranians don't look like Arabs.

I'm all for profiling, but let's do it right. If we start focusing on stereotypical racial characteristics, we'll be overlooking many of those who want to kill us.

There are Caucasian Al Quada members from Iran. Still more simply look Caucasian -- whitebread Kasey Kasem is an Arab, for pete's sake. There are black al Quada members from places like Ethiopia. There are Asian al Quada members from places like Indonesia and the Phillipines. The lily-white, blond Chenians have sympathy for Al Quada

 

[DRZinn]

RE: Carnival Booth

<ahem>

The only logical way to do it (if it made sense to do it at all), given that you can't search everyone, would be to search all Arab males between 18 and 45 (those statistically most likely to be terrorists) AND a random selection of everyone else.

 

[Kurush]

Following the CB algorithm, that proposal works out to be exactly the same as random searching, but with a smaller chance of catching terrorists than you would have had if effort hadn't been wasted on Arabs. Additionally it antagonizes Arab-Americans and Arab countries to no benefit.

 

[TallPine]

Additionally it antagonizes Arab-Americans and Arab countries to no benefit.

Sort of like invading Iraq, huh ....?

 

[Flyboy]

FWIW, a deputy in Palm Beach County, Florida, profiled northbound "drug smugglers" on I-95. His success rate was 86%.

I recognize that he may have been successful, Art, but there's a crucial difference here: hunting drug-runners is a target-rich environment. Even if 100% of his arrests were smugglers, what percentage of smugglers did he catch? I would bet a very tiny fraction.

In the case of would-be suicide bombers, the number of targets is very small--maybe half a dozen or less. Consequently, the odds of nabbing one of them--regardless of selection method--are a whole lot lower than the odds of nabbing a drug-runner. I don't have a good suggestion for how to find them, but using that deputy as an example doesn't work due to the significant difference in circumstances.

Frankly, I'm not impressed with any search-without-cause. I know we have to do something, but I refuse to compromise on my rights. We need to realize that freedom necessarily entails risk, and that we'll just have to accept that risk (we the people; THR members probably already know it). It sucks, but them's the facts. Freedom is a gamble; you pays your money and you takes your chances.

 

[DRZinn]

Following the CB algorithm, that proposal works out to be exactly the same as random searching, but with a smaller chance of catching terrorists than you would have had if effort hadn't been wasted on Arabs.

Actually, as long as people are free to refuse (which hopefully will not change) none of it is any good anyway.

 

[carebear]

I disagree with one key assumption in the Carnival Booth methodology, that both the general US flying population and terrorist populations are "Gaussian".

The pool of transportation-asset terrorists we have experienced (the ones we are screening for) haven't been a random sampling of ethnicities. To throw in scattered individuals, not associated with any organized group and years dispersed, like Tim McVeigh (truck bomber) and the Unibomber (mail bomber) begs the question.

We aren't looking for them because, thus far, they and others like them have not shown themselves to be a threat to the assets we are actually screening for. Likewise, Aum Shiriko didn't attack "Charlie on the MTA", they went after a train in their home territory. Nor have we seen any Indonesians boarding buses in Chicago.

So, realistically, until we get solid evidence that the current crop of ME-based, organized cell, primarily Saudi or Pakistani, suicide bombers have expanded their range of operatives in the US; we CAN in fact add tailored screening to random searches with a net increase in efficiency. The odd Richard Reid would have the same chance of a random stop under such a program as he does now especially with the sniffer technology I hate so much.

Show me some evidence of a 2nd-generation Norwegian grandma joining Hezbollah and I may reconsider.

Although I think the feel-good searches they propose for NY and most of the ones at the airport are more hindrences than help, given their current poor execution.

 

[Rebar]

So, realistically, until we get solid evidence that the current crop of ME-based, organized cell, primarily Saudi or Pakistani, suicide bombers have expanded their range of operatives in the US; we CAN in fact add tailored screening to random searches with a net increase in efficiency.

Although I think the feel-good searches they propose for NY and most of the ones at the airport are more hindrences than help, given their current poor execution.

I agree with both points.

 

[Kurush]

I disagree with one key assumption in the Carnival Booth methodology, that both the general US flying population and terrorist populations are "Gaussian".

A Gaussian distribution is just another term for a normal distribution, a bell curve. If you measure any particular "profile variable" of any random sample of the population, you will get a distribution approximating the distribution in the general population. Assuming a sufficient number of "profile variables" are independent, but not necessarily themselves normally distributed, the sum of the variables will approximate a normal or Gaussian distribution. This is called the Central Limit Theorem. I hope that addressed your concern.

 

[carebear]

I understand what the distribution IS.

I'm just saying that the demonstrated sample pool we are drawing from; ie organized terrorist groups who have within the past, say, decade, have attacked or planned to attack US aircraft or allied mass transit (the pertinant threat), are NOT at this time evenly distributed across racial and or ethnic lines. The guys in 9/11 and the guys in Britain were patently not caucasians of apparent European or North American origin.

They were and apparently continue to be (see London), in fact, of Saudi or Pakistani extraction within a limited age range. Which means we can and could have been profiling based on those characteristics all this time.

With no conclusive or even suggestive evidence of a diversification of the terrorist pool we are specifically addressing, to dismiss profiling based on the supposition of a (apparently non-existent) normal distribution is laughable.

Richard Reid is an exception, McVeigh is irrelevent in motive and target to the discussion at hand. The Czechens and Indonesians have shown neither the real interest or capability to play on this side of the pond and the Iranians haven't shown up at all on the operational side.

Yes, you plan according to capabilities not intentions, thus you randomize; but to not focus on demonstrated repeated capabilities and trends as well is to leave an arrow out of the quiver for no practical reason.

For statistics to work the premises have to be correct and fit the real data sets.