Racial Profiling and Bayes Theorem

People who oppose racial profiling simply do not understand the implications of Bayes theorem. Very few Arabs are terrorists, but almost all terrorists are Arabs. The first point is irrelevant. The second point implies that to maximize your probability of catching a terrorist, your limited screening resources must be directed primarily at Arabs.

"all Arabs are terroists"...give me a break! That cowardly cop, who slammed that sixteen year old kid's head, against that squad car, then popped the kid, with a blind sided punch, is a REAL TERROIST!

The problem with your argument is that because of racial profiling if we just screen arabs, then the terrorists become encouraged to recruit non-arabs. Your Bayes Theory argument would make sense as long as the population of terrorists remains constant. In a real world this won't necessarily be the case.

hear about the fbi agent who shot the kid in the face

I use the term "Arab" loosely to refer to all those nationalities of the Islamic world from which the terrorists are likely find recruits. Sure there are other people who hate America that might be recruited. But right now the large numbers of individuals already trained are of this background, and this will probably always be the case. I can't see large numbers of sympathizers willing to die for the Islamic Jihad coming out of Finland.


Currently in an attempt to appear politically correct, some airports are screening every 10th person in line, even if they are a little old lady from Pasadena. This is a waste of resources which reduces the probability of catching terrorists over profiling, and places our nation at risk. I don't like it one bit.

"but almost all terrorists are Arabs."


How the hell do you figure this?


Ever heard of the IRA (Irish), the ETA (Basque), arguably there are Chechen terrorists, Kurdish terroists (from Turkey's view), Phillipino terrorists, Kashmiri terrorists, the UNITA rebels, the FARC. I could go on. Even of Bush's axis of Evil, only one country (Iraq) is an Arab country.


I'm wondering if you even understand what Arab means?


Paul Talbot

Oh okay,


"I use the term "Arab" loosely to refer to all those nationalities of the Islamic world from which the terrorists are likely find recruits. "


This is of course, preposterous. Arabs make up a small minority of muslims. There are no large muslim countries that are Arab. Egypt is the largest but is dwarfed by Indonesia (the largest), India (over 100 million muslims), Pakistan, Iran and Turkey.


Equating Arab and Muslim is just silly. Of cousre it also ignores the large numbers of Arabs who are Christian (for example about 50% of Arab-Americans are Christian).


Your use of "Arab" is akin to using "Austrian" to refer to all Christians.


Paul Talbot

The terrorists I speak of which are of immediate concern to our airlines as a result of 9/11 are Islamic terrorists. In fact, the 9/11 hijackers were almost all Arab in the correct sense of the word. I think everyone understands who we are concerned about, so let's not muddy the waters with irrelevant technicalities.

That, Mason, is the same nonsense I hear from academics who spend about 5 hours and 5 dollars proving that "technical analysis" of stock price charts doesn't work.


You are saying, basically, that if 1 dollar of effort doesn't yield 10 billion dollars of returns, that it is impossible to get a return, and all effort is useless.


You could just as easily say, well, calling good hands in poker is stupid, because people will learn to read your hand, and then they'll only bet better better ones.


In reality, sound play and effort will raise your win rate by some measurable amount. Even if that rate seems immeasurable, or insignificant, to the casual participant.


eLROY

if you raise every time with an ace showing unless you are rolled up, and in a given situation you do notraise what do you have?


same scenario here. i think you would want to screen arabs at a higher percentage than others, and certainly there is no reason to do overall random screening. but the fact is that you have to screen others for the simple reason that terrorists will simply adjust to a self weighting strategy. saying it should be "primarily" arabs is way too restrictive. maybe a third to a half might be right as a guess.


Pat

you could not be more wrong on this one. we do not want to raise our "win rate" by a measurable amount, we want a strategy that will maximize it. if you think that only screening arabs or muslims will be a good strategy then you are fooling yourself.


Pat

Mason said,


"Bayes Theory argument would make sense as long as the population of terrorists remains constant"


No, even if people mix it up, there are still returns to Bayes Theorem, over just calling every single hand, or every tenth hand, or whatever.


And when you say I "could not be more wrong," you are being uncharacteristically silly, dicaprio. Surely, you acknowledge that your win rate at the end of the day is reduced by the hands you called, or the people whom you searched, unnecessarily.


Every refinement that increases your win rate by a measurable amount moves you towards your maximum win rate.


eLROY

What you are missing, dicaprio, is the cost.


In a poker hand, the reason you have to mix it up, is because their cost to adjust to information is zero. So you have to lower the benefit of their adjusting, by giving them less information.


The cost to Arabs to adjust, by recruiting Tara Reid's grandmother to carry bombs, is astronomical. I say do it, force them to open up their organization to the planet's five blue-eyed granny jihad terrorists.


eLROY

Betting suited connectors costs you money. Not betting aces costs you money. Not screening arabs costs you money. Screening grandmas costs you money.


Folding is free to them. Recruiting blue-eyed grannies costs them a fortune. Screening random people, you're costing yourself a fortune for nothing!


The reason you sometimes slowplay aces, is because your cost is less than their benefit of having information. But they already know you are watching arabs.


AND THEY CAN'T DO ANYTHING ABOUT IT! The amount you cost yourself by screening the first blue-eyed granny is way probably more than the information benefit they lose.


eLROY

eLROY:


There is something known as stratified random sampling. This is where you sample heavier from one group as opposed to another because you have reason to believe that the target population is more present in some particular group. But you don't neglect sampling at least some from all groups. This makes statistical sense and I believe it applies here.


MM

Suppose 1 in 500 thousand arabs is a terrorist, and 1 in 500 million blue-eyed grannies is a terrorist.


Also, assume that your stratified sampling at the gate is by no means the only obstacle to a terrorist.


So, there is no automatic guarantee the granny will get through, even if you state and follow your strategy.


Meaning, they know your strategy, an Arab man, any age, is still 1,000 times more likely to be a terrorist than a granny.


Is your cost for searching an Arab 1/500-thousandth the cost of a terrorist act? Probably not quite.


Is your cost for searching a blue-eyed granny 1/500-millionth the cost of a terrorist act? Probably yes.


Or, adjust it further, to Arabs and grannies who don't get stopped by other means, who make it to the gate.


Is a secretive/closed sect that is forced to recruit foreign grandmothers, more likely to get stopped pre-gate?


You're treating it like a census, like if you sample enough grannies, you will detect a background signal.


Put simply, if the target presence is sufficiently insignificant, there is a cutoff where you sample zero.


On the flip side, you don't want to sample 1 in 10 Arab men, to extrapolate 9 successful terrorists for every one you stopped!


eLROY

Really, random searches are for people who are afraid of getting caught. If you search 1 in 100 people, and their cost of getting caught is 100 times their benefit of getting through, they will be deterred.


Though they may make a bonafide effort, I wouldn't be surprised if some of these folks - like that shoe-bomber Richard Rede guy - aren't indifferent to getting caught. So why random search at all?


eLROY

Timothy McVeigh. All it takes is one person to detonate a bomb. Profiling gives all those who don't fit the profile a free pass to travel the world and do what they wish.

You have two options, SPM. You can either say Timothy McVeigh is profilable, or that the average citizen is a terrorist threat to the USA.


If the average citizen is a terrorist threat to his own country and form of government, maybe it's time to pack it in.


eLROY

You have two options, SPM. You can either say Timothy McVeigh is profilable, or that the average citizen is a terrorist threat to the USA.


If the average citizen is a terrorist threat to his own country and form of government, maybe it's time to pack it in.


eLROY

Maybe we should put all the people opposed to profiling on the flights with all the Arabs we didn't have time to screen. They should have no problem with that, and then everyone would be happy.

which is that profiling doesn't work, involves oversimplifying a complex problem and is ineffective and even dangerous.

I am a blue-eyed caucasian man and, when I recently passed through security at the Albuquerque Sunport, I was randomly selected to be searched, shoes and all. The two men seated next to me on the plane, both of Middle-Eastern decent, had not been subjected to a thorough search. They were speaking Arabic to one another and were receiving many a stink-eye from the caucasian passengers in neighboring rows.


As the plane ascended, the gentleman next to me reached into his bag and pulled out a...deck of cards. There appeared to be no plastic explosives in the deck, and he and his brother proceeded to teach me a card game.


This was too good to be true. Sitting next to two men of Middle-Eastern decent and actually enjoying myself and feeling rather safe?


Well, the man seated next to me again reached into his bag. This time he pulled out a wad of papers. Were they his instructions to hijack the plane and fly it into the Hoover Dam? No, just a bunch of free drink coupons. I drank free Tangeray and tonics all the way to Seattle, courtesy of the Middle-Eastern man seated next to me.


The moral of the story? Should he have been targeted as a potential threat? I think not. Should I have been searched? Yes, a truly random search does not discriminate. What about the "real Americans" who gave the friendly men hateful glares for speaking their native language? They are nothing less than an embarrassment to mankind.

Your alternative - giving nobody a free pass - isn't an option. So, somewhere between nobody and everybody is searching some number of people and not searching some number.


If you roll dice to decide whom to search, and I search the same number of people but Arab males between 15 and 40, my method will "work" to stop more terrorists than yours.


eLROY

The issue with racial profiling is not whether it is effective in catching more criminals. It is clear that it is, and anyone who does not believe that can look at the statistic that drug arrests on the NJ Turnpike were down around 75% in the 6 months after the police stopped their profiling policy. The issue is whether the loss of rights for the innocent members of the profiled groups outweights the value of catching the criminals. In the case of people carring marijuana in their cars for personal use, it is not worth the tradeoff, IMO. However, terrorist attacks are obviously much more serious. The solution lies somewhere between a complete loss of personal rights and ignoring the obvious. Of course, each person's own beliefs regarding the rights of individuals in this country will determine where he stands on this issue. However, dismissing the idea for fear of being called racist is not sensible. My great great aunt Sadie is almost 100 years old, and they searched her and took small sewing scissors from her carry-on bag. If that is not a waste of time, I don't know what is.

we were flying from tampa to new york,and my wife was stopped and searched. she is 32 years old. but the searcher asked her if she was 21 or older. naturally my wife said no and then she didnt mind the search so much. so you see there can be an upside.:-)


pat

I once folded 10-2 offsuit only to find it would have flopped a full house. On another hand a player kept 10-2 offsuit and flopped a full house and got hateful glares from the other players. Should I have played my 10-2? Should the other player have folded it? Should real poker players be pissed off?

The same thing happened on I-75 in Florida. Profiling was extraordinarily effective, but it was stopped. Law enforcement often has as very good read on who is more likely than others to be criminals.


Profiling is not unconstitutional. The constitution protects us from "unreasonable" searches and seziures. The key is what you define to be unreasonable. If a search is part of a larger strategy which greatly reduces the threat of terrorism, then it is not unreasonable. Unreasonable searches would be those conducted for no reason other than to hassle someone that you don't happen to like.


The constitution gives the government the obligation to provide for our defense and protect the general welfare, and that is the only relevant constitutional issue here.

Providing for our defense and protecting the general welfare is certainly a relevant constitutional issue here, but why is it the only relevant issue? Law enforcement might well have a very good read on who is more likely to be a criminal, but they also might be susceptible to racist assumptions and actions and to hassle someone that they don't happen to like.


We decided in World War II that violation of the constitutional rights of Japanese Americans who lived on the west coast was acceptable for our "defense." Yet Senator Inouye was part of a Japanese American military unit that was the most decorated of the war. Maybe racism and hatred obscured our view of which constitutional principles were relevant.


I'm not saying your argument is wrong. I am saying all constitutional principles should be carefully and soberly considered before deciding which are relevant and which are not.

the real constitutional issue here is equal protection. racial profiling is subject to strict scrutiny under the constitution and it fails. profiling is not an unreasonable search and seizure issue because the profiling takes place in deciding who to search. you can profile and still do a reasonable search.


Pat

Actually, profiling does work. It is the abuses of profiling, such as needless harassment, that are to be avoided.

Well IMO that happened because of the idiotic policies currently in place regarding pre-boarding searches.


Anyone who thinks the average blue-eyed Caucasian is as likely to be a terrorist as the average Arab is also either idiotic or has the PC-sunglasses pulled on so tightly that they can't see beyond the nose on their face. Sorry but that's just the way it is.

"You're treating it like a census, like if you sample enough grannies, you will detect a background signal."


That's absolutely correct. When I worked for the US Census Bureau I had to take a course in sampling theory. The idea here is if it is 1-in-500 of one group and 1-in-500 million of another group to do enough sampling of the other group to keep it that way.

You may be right. I'm not saying that this is a great solution. What I am saying is that it should nevertheless be considered despite its problems.

For those too young to remember, the extremist Palestinian terrorist organisations like Abu Nidal's (where is that guy anyway?) in the 70s, were recruiting folks from all over. A Japanese gunman, member of a high-profile Japanese group of the time, was amongst those who in 1972 attacked with machine guns and grenades the passenger terminal of Lod Airport, Israel, killing twenty six civilians and wounding seventy eight others.


Malmuth is correct in his assumption : the composition of your sample will change at any moment.

But the composition of the sample will change only to a degree in this case. The fact is that the philosophical, religious or cultural predisposition to terrorism is far stronger in Arab lands. We could not see, for instance, a sample composition change that would completely invert the current mix. Also I'd bet anything that any group would find it impossible to recruit a grand total of 10 blue-eyed Caucasian grandmothers to be suicide bombers (unless perhaps they were offered enough money to pay for their relatives' cancer treatments, for instance;-)). This sample composition can only change so far.