Game Theory at the Gate

CNN reports that,

The Transportation Security Administration stood by its security officers Sunday after a Florida woman complained that her cancer-stricken, 95-year-old mother was patted down and forced to remove her adult diaper while going through security.

At least from what I can tell from this article, the entirety of the woman’s complaint is that an elderly woman should not be searched so thoroughly at airport security.It sounds like the search was conducted appropriately and in private, and there’s nothing to indicate that officials were rude, forceful, or needlessly degrading.

I remember that this was a standard trope in the post-9/11 debates surrounding racial profiling. Those in favor of profiling would say things like, “It’s absurd to assume that a young Arab man and an elderly Chinese woman are equally likely to be threats.”

While I do believe that there ethical and public policy arguments against racial profiling that go beyond its efficacy as a crime prevention tool, I also believe that short of searching everyone, true randomization is the only unexploitable method of searching passengers. In other words, what if al Qaeda predicts that an elderly woman is less likely to get searched and therefore recruits one to carry a bomb onto a plane? Frankly, I don’t think this would be a bad strategy on their part, and consequently the TSA should not pursue a counter-strategy that could be exploited in this way.

You could argue that even if al Qaeda wanted to pursue such a strategy, there are logistical barriers to their doing so. For instance, they have a much larger pool of young Arab men to recruit from than they do elderly white women. Because so much of their senior leadership is Arabic, they may have an easier time recruiting people similar to themselves. Statistically, an Arab man is more likely to be a Moslem than is an elderly white woman, and al Qaeda is obviously an organization that appeals primarily to Moslems. So, perhaps it makes sense for the US to pursue an exploitable searching strategy simply because they believe their opponents do not have the means to exploit it even if they wanted to.

I see quite a few problems with this line of thinking:

1. Al Qaeda and other Moslem organizations are not the only terrorist threats. The Oklahoma City bombing, for instance, was carried out by White American citizens.

2. Al Qaeda has show some ability to recruit and deploy individuals who do not fit the profile of “brown-skinned male citizen of a Middle Eastern country”. Jose Padilla, accused of trying to smuggle a “dirty bomb” into the US, was a Latino and an American citizen. Richard Reed, the alleged “shoe bomber”, was a British citizen. It isn’t clear that these men were chosen because they didn’t fit the profile, or even that they were chosen at all, but neither of those points is relevant with regard to the exploitability of US counter-terrorism policy.

3. Other opponents in the “airport security game” have been known to attempt to exploit security strategy in this way. Drug smugglers, the original targets of airport racial profiling, have used all sorts of people, from children to old women, to bring drugs into the country. Presumably this is because they were aware of what security officials were looking for and whom they were targeting.

4. It may be possible for an organization such as al Qaeda to employ an unwitting passenger as the carrier of a bomb. There’s a reason an airport official always ask you whether anyone tried to get you to carry anything for them.

I also happen to think that the US is overly concerned about security and wasting huge amounts of money on some of their anti-terrorism measures, but that’s a separate discussion. To the extent that we’re going to employ selective searching as an airport security measure, then randomization is the only unexploitable strategy. If your number is up, then you get searched, no matter what you look like. That means that sometimes Grandma’s diaper is going to need to come off.

Edit: Meant to add that, to the extent that the TSA is pursuing an exploitable strategy that is biased against searching elderly women in wheelchairs, it is great for this case to get a lot of publicity.

27 thoughts on “Game Theory at the Gate”

  1. I don’t see how random searches help at all, since attempts are cheap and the payoff to the terrorists of one success is large. A cursory screening (such as we had before the TSA) is probably about as effective as what we have now. Forcing a strip-search of everyone would be more effective, but in reality the threat is not worth the intrusion and expense…airplanes are not the only target and the fact that nobody has attacked a school or a shopping mall in years indicates that we are not currently the target of any hostilities that would justify even the current level of security.

    The problem is that people can’t wrap their brains around the insignificance of, say, 200 people a year being killed in airplane bombings versus 35,000 a year in automobile accidents.

    • There was just a shopping mall attack early this year in AZ against Congress woman Gabrielle Giffords. Also the number of people killed during 9/11 attack was 2,996.

      http://en.wikipedia.org/wiki/September_11_attacks#Casualties

      Back on topic:
      It does seem like a “keep’em honest” move. Like you’re telling the world “We’re searching everyone.” Wasn’t there a big uproar about a little kid getting patted down also.

      People that think it’s ridiculous to search granny or check the little kids don’t realize the length or the planning that *some* criminals will go through to reach their goal.

      • Congresswoman Giffords was shot by an American left-winger. My comment was about the paucity of attacks on completely unguarded targets by Al Qaeda, to demonstrate that the expense and intrusion of the TSA is not warranted.

        The 9/11 attack was an exploit of our training passengers to be compliant with hijackers. That would not happen again, and it’s likely even the current TSA would not stop the boarding of, say, plastic blades. They are focused on explosions. And molesting children.

        Even if you include everyone murdered in the 9/11 attack, the average number of people killed per year in terrorist attacks in the US or on flights to and from the US in the last 40 years is under 100. The problem is that people get very upset when a plane blows up, even though rationally it’s a tiny risk compared with, say, being fat.

        • Now I’ll admit I follow quiet a bit of news but obviously I’m going to miss some things but when was a school or mall attacked by a terrorist group? Every mall or school shooting that pops to my mind is some “left-wing” extremist or some American that was possibly mentally unstable and/or pushed to their limit.

          I’ve read also that flying is safer relative to say driving. I think a lot of that has to do with how much more strict it is to fly a large plane then say to drive a car. I mean the only think that’s stopping a new 16 year old driver from going 120 miles an hour in the car is well… nothing.

          I’d agree with your last statement but would like to point out that you can do something about being fat. You can’t do so much about a random bomber or hijacker.

          Good discussion btw. All excellent points. 🙂

    • I agree with all of this, particularly the last sentence. My point was only that if you’re going to use selective searches, they ought to be random.

      • *Be* random, or *appear* random? Wouldn’t the best strategy be to scrutinize higher risk threats at a greater rate than low risk threats, assuming that the rate for lower risk threats was still non-zero, and it wasn’t obvious how to put yourself into a lower risk category?

        It seems like even a purely random strategy can be exploited with flood-type attacks–send 5 consecutive people at the same checkpoint, and they won’t all get scrutinized in-depth.

        • Wouldn’t work, if the TSA ever finds an actual terrorist (doubtful) they will take the time to look at other people in the vicinity. You’d be incresing the chances of being thwarted.

          • You’re assuming they’d all act on the same day and out of the same airport. It’s true that when they have any security breach the TSA clears the terminal, so no, you can’t just send five operatives into the same airport all at once. Also, if there’s a successful attack, it will result in far more stringent security measures throughout the country for some period of time afterwards.

            However, Carl is right that increasing the number of attacks is an effective counter-strategy, providing you send your operatives to different airports or spread them out over time.

            For example, if there’s a 20% chance of a lone operative getting caught, there’s only a 4% chance of both of two operatives getting caught. As a bonus, you’ve turned an 80% chance of bringing down EXACTLY one plane into a 96% chance of bringing down AT LEAST one plane, with a 64% chance of bagging two.

            Even so, if you can use only one passenger-selection strategy, random is better than non-random. However, TSA is not limited to one selection strategy. Some of their non-random criteria are immensely stupid and pointless (like flagging everyone who pays cash or buys a one-way ticket), but some aren’t so stupid (like flagging passengers whose name on their photo ID doesn’t quite match the name on their ticket).

  2. Andrew,
    As you point out, there are many issues at stake here, but I’ll keep it on the topic of game theory.

    I think you are correct, but I notice the TSA never uses this reasonable explanation when they are criticized. Either they aren’t employing game theory, so they don’t use it as justification, or they don’t think the explanation is worth promoting. The idea behind an unexploitable strategy is that you can announce it to the world and stll have the strategy work. At least that’s my understanding from reading about the idea as it relates to poker.

    Also, is a purely random search pattern the most optimal here? In poker, aren’t you supposed to bluff/value-bet in proportion to the pot odds you offer your opponent? In this case, shouldn’t we still skew our searches more toward the most likely perpetrators with just enough bluffs (little old lady searches) to keep the opponents unable to guess our strategy?

    Thanks again for giving me something to think about.

    • I can think of a number of reasons why the TSA wouldn’t comment on something like this. For one, they may have a policy of not discussing specific counter-terrorism strategies, though you’re correct that if they were employing an unexploitable strategy, there would be no danger in revealing that information. For political reasons, they may not want to say things like “Even though it may seem like Arabs are the threat…” And then there’s just the question of whether the public, and specifically the ones arguing in favor of racial profiling, would understand the explanation.

      A little old lady with a bomb is not a bluff. As you point out above, the goal is to have a strategy that can’t be exploited even if the opponent knows what we are doing. Announcing “we will search 20% of all Arab males and 2% of all women over the age of 70” suggests a rather obvious exploitive strategy….

      And just to be clear, I don’t believe that the TSA is currently employing an unexploitable strategy. I thikn it’s much closer to the 20%/2% example.

    • when you shop for apples at the market, do you pick your purchases at random? Probably not.

      However, if you’re the apple seller, and you and your family have to eat what you don’t sell, you’d really rather your customers select their purchase at random, as this gives you the best chance (assuming your customers would otherwise pick the better fruit) of winding up with some of the tastier apples at the end of the day.

      The point being that, in most game theory situations, random selection is what you want the other guy to do. It’s almost never what you want to do.

      So, no, random searches are probably not the best way to catch terrorists. That doesn’t mean, of course, that TSA is actually employing the best available methods.

      However, as a deterrent, random searches have a powerful psychological effect. Every time a story about a 5-year-old kid or a 95-year-old granny getting searched makes the headlines, it sends the message to would-be terrorists, who are invariably either very stupid or very insane, that they’re no more likely to slip through than anyone else. As pointed out in some of the other responses, they could defeat this by simply increasing the number of attacks and assuming the risk of getting caught. But they don’t want to get caught, so the random searches do actually make sense as a psy-ops strategy.

  3. I agree with you that profiling based on “objective” criteria is a very exploitable strategy. But how about behavioral profiling, similar to what Israel uses? Basically, they ask everyone a few questions, and based on how things are answered (not the answers, but any “tells” given with the answers), they scrutinize that person more or less carefully.

    That’s a far less exploitable strategy, because exploiting it relies on finding true sociopaths that don’t give off any tells under stress, or using people to smuggle things without their knowledge (which isn’t easy to pull off).

    It’s not much different than questioning your opponent about a large bet, to see how comfortable they are, and modifying your actions appropriately. Few people are really able to answer questions under stress without giving anything off.

    • I’ve heard about the profiling in Israel, but I don’t know much about it, so I’m reluctant to say definitively that it’s good. It certainly sounds more effective than what the TSA does. Even with the need for better trained guards, I’m sure it would be less expensive than installing X-ray cameras in every airport, which is probably why we don’t do it.

    • In Fall 2002 I went on a trip with my mom to Israel and took El Al airlines from JFK. We got pulled out of line when checking our bags, questioned, and got a personal escort from two security officers onto the plane. At the time I looked pretty clean cut, and at first I thought it was weird they were singling us out, but I figured it was because my picture on my passport was from college when I had a long scraggly beard. Then at the end of the vacation when we got to the airport to take our flight back only then did I notice I made a mistake booking the flight and our return flight was to Newark. That must have been a red flag as well.

    • Thanks for the tip, I had no idea. “Moslem” just came out without any real thought behind it.

  5. Random searches are suboptimal, because they are based on the false premise that terrorists want to kill Americans.

    Terrorists do not care how many Americans they kill.

    Al Qaeda wishes strict adherence to an ultra-orthodox brand of Sharia and glorifies jihad as a means to gaining this. Terrorism when it is carried out is either a public relations exercise or an act of war. Carrying out an attack by using a 95 y.o. grandmother would be both unIslamic and inglorious.

    I can’t think of a terrorist group that would derive benefit from a successful attack using a bomb smuggled inside the incontinence pants of a 95 y.o. grandmother.

    None of that is going to change anything though, because random screening does make sense – politically. Americans do fear the terrorists are out to get them by any means possible and like you say random screening is the least exploitable way of preventing this. America is a democracy and therefore caters to this American perspective on terrorism.

  6. The search of granny was almost certainly not random…reading between the lines, I’m guessing she got flagged for travelling on a one-way ticket AND, as a bonus, she set off the metal detectors with her hip replacement/colostomy device/catheter/whatever, which was why they had to take off her diapers, in case it was a Glock setting off the alarm and not one of her numerous prosthetics.

    Which of course has nothing to do with your point about the effectiveness of random searches. Just pointing out that the TSA does do profiling, and one reason you get flagged is flying on a one-way ticket. The obvious counterstrategy to this is, well, obvious.

  7. The problem of terrorism was already solved using game theory more than 10 years ago.
    By Thomas Schelling of USA.
    The key opening question was “why terrorism exists?”
    The final answer:
    “If you want to avoid acts of terrorism carried out against people in your country, avoid getting involved in the affairs of other countries.”
    The alternative: dystopian society

  8. “Congresswoman Giffords was shot by an American left-winger”

    I don’t know if this was a deliberate slur or an honest mistake but that’s categorically untrue. Even the most cursory look at Loughner’s deranged ramblings will show that whatever he was he wasn’t left wing – he held many views utterly at odds with left wing philosophies, and which no one claiming to be left wing could consistently hold. His overall philosophy seems to have been very much an anti-government one. It would thus be much easier to make a case for him being on the extreme right than the extreme left. But trying to put him into a particular bracket like that seems like little more than petty point-scoring in any case – he was (sorry, is) a depraved, dangerous lunatic, and dissecting his worldview to see what side of the spectrum we can put him on doesn’t seem like a worthwhile or fruitful exercise.

Comments are closed.