racial profiling

[JBA]

The Religion of peace

1972       Munich Germany
1979       Embassy In Iran
1980       Lebanon kidnapping and murder
1983       Beirut bombing
1984       Achille Lauro 70 Jewish American in wheelchair thrown from ship
1985 T     WA 487 hijacked from Athens
1988   Pan Am flight 103
1993    WTC
1995    Kabul Towers in Saudi Arabia
1998   Embassies: Sudan, Tanzania
2000   USS Cole
2001        WTC, Pentagon, Penn. Headed for Washington
2001    Shoe bomber plane
2001    Decapitation reporter Pakistan
2001   LAX airport shooting
2001        French tanker bombing
2002   Bali night club
2002    Snippier shooting USA Washington
2002        Moscow theater
2002   Business man in Jordan

[capt. apathy]

profiling is just another word for paying atention to trends.
most of us do it at work.  but since it's not 'racial' profiling it's ok.

for example.  I have a '89 jeep cherokee with the GM straight-6.  I had a exaust leak. so I take it to the machanic.  I tell him I need an exaust leak found and fixed on my rig.  he asks what I'm driving, so I tell him. without getting out of the chair, he tells me I have a cracked exaust manifold and quotes me the price to fix it right off the top of his head.  I was kinda taken by suprise and asked how he knew the price so quick.  he says the stock manifolds crack all the time and need to be replaced with the non-cast type. he says just about all of these end up with a cracked manifold and if you replace it with stock it happens again.

the point is that it is not an unusual thing.  do jeeps have exaust leaks that are not due to a cracked manifold? yes. do other brands of cars get cracked manifolds? yes.  but by paying attention to different problems and where they often come from he was able to diagnose my car with alot less effort, time, or money.  he knew his job and went and checked the manifold first,  saving alot of time looking around the car.

I see nothing wrong with taking a closer look at people who fit the discription of the guys who bring most of the problem.

[Thrawn]

Originally posted by Curval
The "might not be" part of the above sentance intrigues me.

It implies the possibility that it MIGHT BE the best way to stop actual terrorism.

Of course it might be, I also might win the lottery.  No one has countered the Carnival Booth Algorithm paper yet, which indicates that it is not the best system.

JBA,

Religions don't kill people.  People kill people.



Capt,

How would your analogy change if the mechanic was actively trying to defeat your profiling system?

[SOB]

Your mechanic is obviously a racist.


SOB

[Seeker]

What a disturbingly anti-semitic thread.

[Dowding]

Anti-semitic?

[midnight Target]

Sounds like the terrorists have almost won the war.

[Gunthr]

Thrawn said: "It has been proven that racial profiling is a faulty system for security as it leaves gaps that can be exploited. Random searches are more effective."

I asked for your source. You gave:

"Carnival Booth: An Algorithm for Defeating the Computer-Assisted Passenger Screening System

http://swissnet.ai.mit.edu/6805/stu...papers/caps.htm"

Thrawn, your source does not "prove" that profiling is ineffective.

Your source is a paper prepared by students for a project. It is regarding the FAA's use of a specific Computer Aided Profiling system in 1999 for airport passenger screening. The FAA then, and the Homeland Security agency now do not rely on that alone. It was a system in development back then anyway.

As pointed out elsewhere, profiling is never perfect, but it really does work.

[Curval]

Originally posted by Dowding
Anti-semitic?

Sure..Arabs AND Jews are semites...surprised you didn't know that Dowding.

Funny me being called an anti-semite, but there you have it.

[capt. apathy]

quote
_______________
Capt,

How would your analogy change if the mechanic was actively trying to defeat your profiling system?
___________________

I don't follow your point.  are you saying the airport security people are activly trying to screw up any usefull profiling?

because if you pay attention and follow the analogy.  the security people would be the machanic (searching for the problem).

[miko2d]

Thrawn: Making people sit on the back of a bus because of race is racism.
Making people sit on a different plane because of race is racism.

 I would agree with that. That is of course quite separate from my support of freedom of any private individual/business to practice racism, however disgusted I am with such individuals. Public companies and businesses operating under the public license should of course be disallowed from precticing racism.

 But just listing cases is not enough if you can freely add to that list. Like, this whole thread is about airport searches and you did not list it. I was hoping for some more general principle.

That would be for experts to decide. "
That's nice.  What problem do you have wtih paper I linked to?

 I did not mean specific experts who's opinion becomes a digma, but experts entrusted with each particular thing. Experts views differ. Experts views change.
 The linked study adresses possibly exploitable shortcomings of a particular computreised subcomponent of a security system. That is why such system are going to be improved and work in conjunction with human experts.


Eagler: it ain't about STOPPING terrorism (as this is impossible), its about MINIMIZING it.

 You should add "in a cost-effective manner". We could eliminate 100% of traffic accidents tomorrow if we wanted. It's just that most of the population would die of desease and starvation within a year.

 One of the most brilliant minds of our generation - economist Thomas Sowell, who is black, by the way, has a good book on the cost of knowlege in market economy - "Knowledge and Decisions".

 It may serve as a good justification for many "rasist" practices based on the fact that obtaining particular knowlege is expencive and using general knowlege (about race/ethnicity) really helps cut down on the costs. Of course the optimum is a mix of certain proportions.

 We successfully use encoding to compress files and save (money, bandwidth, storage) based on our knowlege of distribution of letters. Statement "the next letter is 'a'" contains much less information than "the next letter is 'z'" - because a is more common. That's why we can encode 'a' with 2 bits and waste 9 bits on 'z' and still save space - which we can use to fit in more letters into the same space. How much bits to spend per letter is determined scientifically.

 Statement "this 'white person' is not a terrorist" contains much less information that the statement "this xyz is not a terrorist" because xyz are more likely to be terrorists according to the statistics. Spending 10 minutes in order to verify that is much more of a waste than spending 10 minutes in order to verify that xyz is not a terrorist.
 So the queston becomes:
 What strategy/mix we employ to get the highest probability that there are no terrorists among N people aboard a plane which contains a fraction M xyz with fixed limited number of resources (searches)? The random search is certainly not the optimal solution. It's pure and relatively simple math and has nothing to do with racism. It must lead to the search of some 'white' people but not proportional to their numarical representation.

 Again, that is purely compurerised algorithm, subject to algorithms intended to break it's operation - that's why random element and human factor will always be present.

 miko

[Thrawn]

Originally posted by capt. apathy
I don't follow your point.  are you saying the airport security people are activly trying to screw up any usefull profiling?

I'm saying that Al-Quaeda is activly trying to exploit security problems with profiling.

Gunthr,


Can you give me some ideas on how this solution:  (1) Probe the system by sending an operative on a flight. The operative has no intent of causing harm. He has no explosives. He has no weapons. He has nothing. He simply takes the flight and notes whether or not CAPS flags him. (2) If he is flagged, then send another operative in the same manner. (3) Repeat this process until a member who consistently eludes CAPS flags is found. (4) Now send this operative on a mission with intent to harm, complete with weapons or explosives. Since CAPS didn’t flag him last time, he likely won’t be flagged this time, so he incurs much less risk of special scrutiny., would not not defeat a profiling based security system?



Miko, thanks for the response.

[Thrawn]

miko, what happens when someone decodes your algorithm?

[miko2d]

I'd like to remind you again - the goal of a system is not to prevent possibility of terrorism, but to make it not cost-effective.

 Terrorists may come up with the optimal way to act. That way will not be 100% effctive and it may take them a lot of money and many years and exposure - during which their operations can be compromised - even if the algorithm they plan to defeat does not change, which is a stretch.
 So they might try some other way - probably less hurtfull.

 Likewise, the terrorists do not have to succeed with every act. They can come up with a strategy causing us a lot of loss with very few actual direct damage.

 It's a facinating minimisation problem. Of course the expectation that our legislature does not operate on scientific principles and waste resurces trying for impossible "comprehencive" solutions must be part of their plan.

 miko

[miko2d]

Thrawn: miko, what happens when someone decodes your algorithm?

 I am in a business of computerised stock trading. I know for sure that there are ways to make money on algorithms - which work against competitor's algorithms as well as "spontaneous" market dynamics. Decoding competitors' algorithms is part of what we do. Algorithms can be relied upon. But you do not have to sell me on a notion that any fixed algorithm can be eventually defeated.

 The same algorithm is never good for longer than a few weeks/months. Market changes in responce to algorithms used. So one has to adapt. And one cannot avoid using live humans with their brains and intuition.

 Anyway, this whole thread subject was originally human-vs-human affair. The algorithm cannot be rasist and an algorithm saying "skip all whites, check all arabs" can hardly be called an algorithm or need a computer to operate.
 The issue was that humans "unfairly" subjected other humans to examination. They were certainly somewhat computer-assisted but the decision was theirs.
 I was never arguing in favor of any algorithm, however successfull at the moment - just that "preferential" (not exclosive) treatment of certain identifiable categories of people is justified and does not violate civil rights.

 miko