He Bought 6000 Rounds

OK. Seriously. So have I and I’m not a serious gun guy. 6000 rounds isn’t that exceptional a quantity. I did buy .22 LR which isn’t going to be the first choice of for your garden variety killing spree. When I go to the range, I typically shoot 200-300 rounds. If you go shooting once or twice a week, buying in quantity is what you do. If you fire larger calibers than .22 LR … you collect brass and do you own reloads (9mm for example costs about 40 cents a round if you don’t do reloads). It takes repetition to teach your spine to shoot.

Question for the gun controls must be higher and higher? If 50 of the people in the theater were armed via concealed carry … Might the outcome have been different?

The quoted link above clearly thinks that tighter gun laws are necessary, yet the recent shooter had a large cache of illegal explosives. Clearly gun laws weren’t what stopped him. The shooter had no history of mental illness, crime records, or anything to prevent him from buying guns and moderate stores of ammo in any state. His emphatic “make it tougher for criminals and nuts to buy guns” would have done nothing. So, then why bring it up now?

Recently, “America is no longer ‘top nation'” inspired by insipid political advertisements in the guise of Allen Sorkin TV-News drama arose in conversation. This was a big lie cementing a reasonable argument (that we have problems that need addressing). Basing an argument on a lie, however good the argument is, weakens your rhetoric and destroys your credibility.

Look people, both right and left, but it seems lately the left of the aisle turns to this more and more (see the above and AGW) … stop it!! Stop using clearly stupid/false things to persuade. If you think tighter gun laws will help bring down gun crime. Prove it with actual data. If you think on the other side, that armed homeowners and women with concealed carry are safer on the streets at night against predators) use actual data not specious easily disprovable remarks. Against both sides of the gun control argument, data I’ve seen has shown that looser gun laws and concealed carry is uncorrelated with increased gun violence. One study demonstrated that legalized “open carry in bars” had no impact on gun violence in a city (or was it cities). Seems to me if restricting or not restricting personal liberties has no impact either way on our society … then “more personal freedom” should be the default choice.

Leave a Reply

Your email address will not be published. Required fields are marked *

34 comments

  1. Boonton says:

    After 9/11 there was a brief scare in NJ about some Arab appearing men who were showing up at Shop-rites buying shopping carts full of baby formula. The question that was immediately raised was this part of some new terrorist attack on baby food.

    Turned out they owned bodegas in cities like Newark and Paterson. Given the price they sell baby formula and the deep discounts that sometimes show up as ‘loss leader’ sales out here in the sticks it turned out it was more profitable for them to buy the baby food in bulk retail rather than wholesale. Lesson I take from this is that a lot of things probably happen in our vaste economy that would look pretty strange until you got to know the dynamics more intimately.

    So I agree, ‘6,000 rounds of ammo’ tells me next to nothing. At the very least I’d want to know what the average ammo purchase is, what the standard deviation is. Such data is probably not publically available to either me or even law enforcement. I suspect there are plenty of very small businesses like shooting ranges and such, such businesses might show up in ‘big data’ as individuals making large bulk purchases.

    Let’s even say that an individual buying 6K rounds in a single transaction is highly unusual. Let’s say of the set of people who are not registered as gun dealers or traders, there’s only 1,000 such transactions per year. Well that sounds great except given the last 20 years we’d have 20,000 transactions but only 1 incident. That means if we ‘profiled’ such transactions it would take 20,000 times to catch one person. (And what does ‘profile’ mean here? How would one approach this guy *before* he did anything?)

    I’m going to flip this one on Mark. He feels profiling can substitute for universal screening in airports so why not do it for all gun transactions? Profiling works best with data and terrorist actions are so rare that there just isn’t a large population to build a profile that’s very reliable. But criminal activity is much more common and it should be possible to build profiles of gun transactions that rarely end up in criminal activity (say the stable person whose been doing guns as a hobby for 30 years, never in any trouble with the law) versus ones that do (say the one guy the Fast and Furious operation was tracking who was on food stamps yet was spending in excess of $100,000 buying heavy duty guns at auctions in the US).

    It seems to be consistent he should support profiling with guns and ammos. Granted it may not be as simple as ‘examine anyone who buys 6,000 or more in a single purchase’ but that’s just a first take at the data.

  2. Boonton says:

    The other thing to consider is that it’s wickedly easy to hack profiling systems. A few years ago when we got a dog and registered him (since he is a pure breed Irish Setter), it wasn’t very long before we started getting lots of ads for various pet related products. A while ago I was looking at pool filters, now whenever I go on some sites I’m shown adds for pool filters. Worst of all was when I innocently started filling out an inquiry for refinancing my mortgage and stopped halfway through, I’m still getting 3-4 calls a day from that.

    But I could do the reverse. I could own a bird and buy all the supplies with cash and never show up’ on the grid’ as someone likely to own a bird. I could make myself appear to be an avid collector of lawn gnomes, or guns, or a skydiving enthusiast, without ever actually being one. Marketing profiling only works because most people do not care enough to put even a halfway serious effort into either hiding their interests or creating fake interests.

    But a terrorist would and unless you’re going to establish a big brother state where nearly everyone has their own FBI ‘data file’ it’s always going to be wickedly easy to ensare the wrong people in a profile while letting the halfway clever person dodge it. For example, if there was a profile of people buying 6,000 rounds, a kid this smart could have easily spaced out his purchases or purchased from several different sources or enlisted other people (real or faked) to amass his stockpile. Consider pseudophed, which you need to show your driver’s license to buy now because of meth. I’m sure there’s a lot of really dumb people who use very clever ways to dodge that control.

  3. The difference between you and I is that I’m for empiricism on all issues, regardless of political alignment. When the empirical evidence demonstrates that my side is wrong on an issue (as I agree it does on this one!) I go with the evidence.

    You? You cite empiricism only when convenient. In other arenas, you think:

    * That 2,000 year-old third-hand testimony is absolutely convincing.
    * That personal anecdotes are more important than data.
    * That polls are “cricket races.”
    * That “common sense” is more important than data.
    * That data can be ignored because scientists are biased.

    Etc.

  4. Also, profiling is dumb. Read this debate between Sam Harris and security expert Bruce Schneier.

    Note that I was biased towards Sam Harris, atheist luminary that he is, but I think Schneier demolished him. Profiling isn’t bad because it’s immoral, it’s bad because it doesn’t work.

  5. Boonton says:

    I’ll hold off reading the debate until tonight but I’ll say that the question of profiling working or not depends on what you’re trying to accomplish.

    Are you trying to fish? In other words catch a few members of a larger population? Then profiling will often work. When Amazon profiles and estimates that people who buy a Rush Limbaugh book will be more likely to also buy a Sean Hannity book than the general population, profiling can and does work.

    But if you’re trying to capture *all* of the population it won’t. If you want to capture everyone who might be willing to buy a Sean Hannity book, profiling will fail you. And unlike Hannity customers, the hurdles for profiling to succeed go up once you model profilees who respond to profiling by trying to thwart it, even if they are only halfway intelligent in doing so.

    The problem one faces with terrorism or mass killings like this is that ‘fishing’ is not an acceptable solution. Amazon is perfectly happy to loose the potential eccentric person who buys a lot of books by Marx but would buy a Hannity book if prompted if it could capture a lot of other Hannity sales. In general we want to be 100% free of terrorist attacks, not simply capture/deter a portion of them.

  6. Yes, obviously profiling has uses in OTHER arenas.

    You correctly mention one very big problem using profiling for terrorism, but perhaps an even bigger one is that, unlike Amazon customers, the terrorists are trying to fool the system. Imagine how successful Amazon’s recommendations would be if you were deliberately trying to throw off the algorithms.

  7. Boonton says:

    I would say if you have a lot of data, you might develop a useful system of profiling. For example, there’s a lot of credit card fraud so it wouldn’t be surprising if there’s good profiles of attempted fraud that ‘work’ (meaning catch a lot of it, not catch all of it). But I think there’s two problems:

    1. Not enough terrorists to build a profile
    For example, I suspect one might develop a profile of car bombers at checkpoints in Iraq or Afghanistan. But that’s pretty localized terrorism. How many 9/11 hijackers or UK/Spain train bombers or Tim McVeighs do we have to build a database of the high profile ‘celebrity terrorist incident’? Reality is we probably couldn’t even get together 100 people to enter into this database. We are looking for a profile to describe a very, very thin part of the tail. This isn’t like Amazon predicting sales of the next Sean Hannity book, it’s like Amazon trying to predict if some unpublished, new author will make 100 sales or 1000.

    2. Not enough terrorists to test a profile

    Say you claim to have a profile that can pick up another 9/11 type event. Given a pace of maybe one such thing attempted every 3-5 years how are you going to establish that your profile works? In a previous post Mark linked to a blogger who claimed the US system of selectively scanning 1% of incoming cargo was a great way to combine profiling with a wise allocation of resources. But since we know of no attempt to smuggle a WMD into the US via the cargo system we have no idea if the profiling system works or if it just happens to be getting credit for the fact that no one has tried to use it for a major terrorist attack.

  8. Mark says:

    Boonton,
    Back up a minute. What is is the purpose of our profiling? To speed and lower the expense of travel in airports. Before 9/11, business travelers would regularly arrive 30 minutes prior to departure to catch a flight, 45 minutes at the outside. Now 90 minutes is more the norm, and 45 minutes is cutting it short. So, you have billions of wage-hours lost to TSA besides the search expenses themselves. So there is are big money reasons to smarten the search. If you take three “levels” of search, cursory (pre-9/11 walk through metal detector), to standard TSA, to the current heightened search that attractive but not recognizably famous women recieve (pat down + x-ray + nitrogen). So to save the money a majority of the passengers would expect a lighter search and those with the higher search levels would have smaller lines for their search wait.

    So how to make a profile that works? Use publicly available information against an unknown trigger algorithm which changes all the time. What is the cost of a false positive … a bump up the search intensity. So, you’re a terrorist that wants to try to game it. 10 of your buddies try sneaking stuff in a variety of ways. Some get caught, which means that people connected to them automatically get bumped. Oops. Change of habits, of any sort. Bump. Oops.

    JA,

    Imagine how successful Amazon’s recommendations would be if you were deliberately trying to throw off the algorithms.

    Exactly how do you explore the algorithms? Hmm? Unlike Amazon which publishes clearly (for obvious reasons) the results of their algorithms every time. You don’t know why, how close, or if you got bumped for a reason or by random. Do you think a clever fellow (such as yourself playing devils advocate for more than 30 seconds) can figure out an algorithm that’s hard to decipher/figure out? Thought you could.

    You cite empiricism only when convenient. In other arenas, you think:

    * That 2,000 year-old third-hand testimony is absolutely convincing.
    * That personal anecdotes are more important than data.
    * That polls are “cricket races.”
    * That “common sense” is more important than data.
    * That data can be ignored because scientists are biased.

    Etc.

    Let’s see.

    * I don’t base my religion solely on 2k old third hand testimony.
    * personal anecdotes and “common sense” are not more important than data. However, I went to school at a place were a fellow named Enrico Fermi taught and is thought highly of. Have you heard of him? He (and Feynman) teach that you that a theory has to be reasonable, that things that are reasonable can be roughly calculated by back-of-the-envelope means, and to use your own mind. That would be the common sense thing. If you want to figure, for example, that a few percent change in a parts-per-million atmospheric gas whose abosorption line is already saturated is going to make big change …. tell me why. Tell me how. Lay out some equations that make your argument instead of citing political documents like you did that last time asked.
    * “That data can be ignored because scientists are biased.” Huh? You do that all the time.

    Have you read Polanyi yet?

  9. Mark says:

    Boonton,
    Do you think a few guys with box cutters could still hijack a plane? Would it take a bit more, seeing that the hijack game has changes rules.

  10. Boonton says:

    Back up a minute. What is is the purpose of our profiling? To speed and lower the expense of travel in airports.

    The purpose of screening is to achieve zero terrorist attacks on the air system. Profiling, by definition, seeks to achieve saved time and money by relaxing that goal. Pre-9/11 terrorist activity was tolerated to some level on airlines. Quite a few cases of airline terrorism involved no casualities or very limited ones. The absolute worst cases entailed everyone on the plane being killed. 9/11 demonstrated that it could be much worse than that therefore zero tolerance was decided as a priority.

    So to save the money a majority of the passengers would expect a lighter search and those with the higher search levels would have smaller lines for their search wait.

    Actually the majority of passengers do go through a light search if by that you mean walking through a scanner. Since we have the tech to do better than a simple metal detector why not do that?

    So how to make a profile that works? Use publicly available information against an unknown trigger algorithm which changes all the time.

    And how do you measure how good that algorithm is? If you get a very good algorithm, why would you want to change it? How exactly are you changing the algorithm if you’re *NOT* trying to match public information on known terrorists?

    Also, keep in mind, you have an over the top view of what is publically available information. For example, you seemed to think a person’s Amazon purchases and Google searches are publically available information, they aren’t.

    And, of course, we come back to the question of what information was known about this shooter that would have triggered a profiling algorithm?

    Do you think a clever fellow (such as yourself playing devils advocate for more than 30 seconds) can figure out an algorithm that’s hard to decipher/figure out? Thought you could.

    If the algorithm is intelligent it is either based on information from previous terrorists or it is based on what someone thinks a terrorist would look like. If its the former then a terrorist organization could game the algorithm by using agents with radically different characteristics than previous terrorists. If its the latter then they would again estimate what they think the CIA thinks a terrorist looks like and work that way.

    If it’s based on neither then all you’re essentially doing is a random portion of passengers. If you do that then you have two problems:

    1. The ‘sweet old lady’ getting randomly selected for intense search while 20 something Arab guy from Pakistan doesn’t.

    2. If the % is set very low at a certain point the terrorist will just take the chance. If you’re only going to randomly screen 1% of passengers, then you have a 99% chance of getting through. If the % is set very high, you might as well stay with the universal screening coupled with random and profiled ‘intenstive’ searches.

    Do you think a few guys with box cutters could still hijack a plane? Would it take a bit more, seeing that the hijack game has changes rules.

    Indeed it has, but as we can see the terrorists have changed their game too. The rise of the ‘underwear bomb’ attempts indicates that they know another 9/11 would require more than boxcutters.

  11. Do you think a clever fellow (such as yourself playing devils advocate for more than 30 seconds) can figure out an algorithm that’s hard to decipher/figure out?

    The problem is that the algorithm has to work, too. Therefore, there’s only so much you can do to obfuscate it. There are certain obvious countermeasures terrorists could take that would make a system dependent on profiling worse than a system without profiling — recruiting non-Arabs, old ladies, children (because they agree with the cause or by coercion), buying round-trip tickets, etc. Profiling therefore adds cost, complexity, and risk to the terrorist cause which is a very good thing, but it also opens up some vulnerabilities such that when terrorists can get over those hurdles, their chance of success is drastically improved, so overall it’s a bad thing.

  12. Boonton says:

    Note JA’s semi-profile is fine for only a very specific type of terrorism, Al Qaeda inspired 9/11 style terrorism. What about pure Rampages that people like John Holmes just went on? Since he is not an Al Qaeda terrorist and he doesn’t fit the profile of one he would be green lighted for easy entry onboard a plane.

    Now how do you modify your profile to catch both Al Qaeda terrorists and mid-American Rampagers? Mark suggested his recent purchase of 6,000 rounds would do, but then he claimed that’s not such a strange purchase. Perhaps his explosive purchases but why do I suspect they happened off the big data grid?

  13. Mark says:

    JA,

    The problem is that the algorithm has to work, too. Therefore, there’s only so much you can do to obfuscate it.

    OK. That was 5 seconds. Keep going. Can you fool those countermeasures?

    Boonton,
    JA’s semi-profile is bad. Give him a chance. He’s only used 5 of his 30 seconds of thinking. And no I didn’t suggest 6k rounds is the trigger, but his recent foray into guns ownership. I suggested any change in behavior patterns as suggested by purchases bump up your security scan level.

  14. OK. That was 5 seconds. Keep going. Can you fool those countermeasures?

    Not if I want profiling to work! At some point your profile consists of “everybody.”

  15. I don’t think there is a good solution to stopping massacres like the most recent one. It’s a very low-probability event that’s probably just going to happen regardless of what we do.

  16. Boonton says:

    And no I didn’t suggest 6k rounds is the trigger, but his recent foray into guns ownership.

    Why not recently dying his hair red (or orange)? Would someone who does neon blue clear the profile because to date we’ve had no oddball rampages from blue haired nutcases?

    More importantly you’re neglecting the problems. Let me give you a briefer summary:

    1. No way to measure whether your profile is accurate.

    2. If you’re randomly changing your profile, then you’re saying accuracy isn’t important. So then just do random screening.

    3. You’re vastly overestimating the ease of collecting and properly reading this info.

  17. Mark says:

    JA,
    I think we’re talking about different things. We’re not talking about profiling to prevent low probability events such as the theater killings, but to prevent same said person from doing that on an airplane.

    Not if I want profiling to work! At some point your profile consists of “everybody.”

    You want to lower the costs appreciably with a small impact on effectiveness. Unless you pretend that the screening has no costs … which leaves no room for discussion … just strip search everyone. There’s a cost/benefit analysis here. My original discussion with Boonton were premised on the observation that there is no information of subject given to TSA screeners, and using that could make the search process both faster and more effective.

    Boonton,

    Why not recently dying his hair red (or orange)?

    If you can detect that with your scan, sure. And especially if the bad guys start thinking that’s a possible trigger for more careful scanning. Just got married, bump to medium. Changed churches. Bump. As for ease of collection, review the security post in today’s link/post. And the randomness (or quasi-regular randomness) is just there to mask the real algorithm. Random screening isn’t as interesting as non-random, but changing regular patterns … you’d keep just not quite getting the trigger/pattern.

  18. The other issue is who is executing the profile. If it’s your average TSA agent, your profile needs to be simple enough for someone with an average to below-average IQ to execute.

    This is the big difference between American security and the Israeli security that people like to hold up as an example. Israel is not only small, but spends money to hire intelligent, competent people to use their own intelligence and judgment to make decisions. I don’t think we can count on TSA agents to do anything like that.

    Part of this has to do with scale and part has to do with culture.

  19. We’re not talking about profiling to prevent low probability events such as the theater killings, but to prevent same said person from doing that on an airplane.

    It’s also very low-probability (for each flight) on an airplane.

    You want to lower the costs appreciably with a small impact on effectiveness. Unless you pretend that the screening has no costs …

    Since I think effectiveness is already so low, we could lower the costs almost to zero without much impact. That’s another reason discussion of profiling is so silly. You’re looking for a way to optimize a system that already doesn’t provide much security.

  20. Mark says:

    JA,
    You’ve said that quite often (“doesn’t provide much security”). Yet you’ve never substantiated that or supported your claim. Wanna do it now?

  21. Mark says:

    JA,

    The other issue is who is executing the profile.

    Since the profile’s we’re talking about are all background check related, seems to me that would be automated. Scan the ID, which indicates what sort of check to do.

  22. You’ve said that quite often (“doesn’t provide much security”). Yet you’ve never substantiated that or supported your claim. Wanna do it now?

    I have in the past provided links showing how easy it is to sneak items past security.

    Since the profile’s we’re talking about are all background check related, seems to me that would be automated. Scan the ID, which indicates what sort of check to do.

    Sure, I agree that there are some things we could do with that method that would be beneficial. That’s not really what I’m arguing against. I’m arguing against profiling based on race, religion, gender, age, etc.

  23. Mark says:

    JA,
    I don’t recall any links. I recall you and Boonton discussing how easy it was to sneak stuff by and that you thought it fairly trivial to bypass TSA security. I don’t recall any substantiation of the claim.

    I’m arguing against profiling based on race, religion, gender, age, etc.

    OK. I’ve never advocated stupid/trivial algorithms. Now you’re up to 10 seconds of thinking about methods of using information … and you’ve found something perhaps useful. Wanna go for a whole 20 seconds more? Suggest some methods of using information that would be useful and hard to beat by gaming.

  24. Boonton says:

    We’re not talking about profiling to prevent low probability events such as the theater killings, but to prevent same said person from doing that on an airplane.

    This is incoherent, a plane attack is also a low probability event. If your profile couldn’t prevent a theater killing then it can’t prevent a plane one making it useless.

    In terms of lower costs/saved time. Here is the bar your profile must clear:

    What is the probability that your profile will capture a wannabe killer/terrorist? Let’s say you have a very fuzzy profile that will flag 20% of everyone. That would work if in that 20% you get 100% of the terrorists. Yes you’d still have millions of searches, but you could save time and money on 80% of everyone else.

    But if your profile only has a 20% chance of capturing, then it’s no better than just randomly doing 20% of everyone. But since we’ve already established we want zero successful attacks on planes, 20% is not acceptable.

    So how can you establish that you have a profile that will clear a large amount of innocent people but be almost perfect catching bad people?

    My original discussion with Boonton were premised on the observation that there is no information of subject given to TSA screeners, and using that could make the search process both faster and more effective.

    This is a false premise. The status quo is everything. Everyone is screened, there’s random intensive screening, there’s profiled intensive screening (i.e. ‘no fly lists’) and there’s also room for TSA agents to more intensively scan ‘based on their hunches’. While measuring effectiveness is difficult, we have good reason to think this could achieve near 100% preventing of people beinging serious weapons (guns, large knives, conventional explosives) on board planes.

    You propose to drop one of the legs (universal screening) in exchange for beefed up profile screening. You’re not going to make a case for more effectiveness as you can’t improve much on near 100%. You can only make your case on lower cost without the loss of effectiveness. IMO you haven’t even begun to map out a convincing argument.

    If you can detect that with your scan, sure. And especially if the bad guys start thinking that’s a possible trigger for more careful scanning. Just got married, bump to medium. Changed churches. Bump. As for ease of collection, review the security post in today’s link/post.

    You have a hopelessly silly view of modern databases. Believe it or not there is no single ‘marriage database’ in the US, which is why it is still possible for men to sometimes get caught committing bigamy. As for ‘changing churches’, please tell me how is the TSA supposed to find out that you recently changed your church?

    Random screening isn’t as interesting as non-random, but changing regular patterns … you’d keep just not quite getting the trigger/pattern.

    It’s interesting that you think this is effectively any different from randomness. You’re basically saying you’re not even trying to get a good profile, because even if you did it would end up getting changed every few days! So it is effectively the same thing as a random screening. In the case of the random screening, the terrorist has to calculate his odds of being pulled into the screening. If 1% of passengers are, then his odds are pretty good, if 75% then not so good. From your side, though, randomly screening a large percentage of passengers does not address your goal of reducing costs and time.

    You’re at an impass here, either go with a good profile system or go home.

    JA

    I have in the past provided links showing how easy it is to sneak items past security.

    some items maybe, but to be honest I wouldn’t want to try to sneak a gun aboard a plane. Terrorists seem to also disagree with you as they are opting to try to develop high tech solutions to dodging security rather than trusting your links that it’s easy to get stuff on board planes.

    IMO the problem with Mark’s proposal is that it drastically underestimates the costs with even beginning to try to build a reasonable profiling system, the beginning of which would mean having the gov’t build huge centralized databases on each and every person it can find, making J Edgar Hoover’s infamous files look like baby legos and making the old X-files series writers blush.

    A better solution is to just perfect the scanning option. Everyone walks through a scanner that not only picks up metal but can pick up explosives and ‘odd things’ (like a rectangle with wires and batteries taped to someone’s back) for detailed examination by TSA agents. For most people it’s no different than walking through a metal detector. The whole ‘privacy’ issue can be addressed by having pics reviewed by AI or by analysts away from the airport. Granted someone with the backing of a Mossad or KGB or CIA or James Bond’s support team may be able to counter high tech with higher tech, but then they could probably game a profiling system even easier.

  25. Mark says:

    Boonton,

    This is incoherent, a plane attack is also a low probability event. If your profile couldn’t prevent a theater killing then it can’t prevent a plane one making it useless.

    Count the number of airports in your state. Then compare to the number of malls, stadia, and theaters. Count the number of exits and entrances for each. You think this is the same problem?

    But since we’ve already established we want zero successful attacks on planes, 20% is not acceptable.

    At any cost. Riiight. You’ve left the discussion. Let me know when you want to come back. There is a cost. There is a benefit. Therefore there is a cost benefit curve, the question then becomes what matters yeild the same effectiveness with lower costs and where do you want to set the effectiveness. You note a “no-fly” list … but apparently it is far far better to have exactly one criteria … question why can’t a domestic no-fly individual submit to a careful search and then fly? Hmmm? Why can’t there be more gradients between no-fly and several search criteria.

    You’re at an impass here, either go with a good profile system or go home.

    Well, in 20 seconds of work devils advocacy (or similar from you) the beginnings of how to structure a working good profile that can bend down the cost curve will be imagined. So, you can concede that point.

    IMO the problem with Mark’s proposal is that it drastically underestimates the costs with even beginning to try to build a reasonable profiling system, the beginning of which would mean having the gov’t build huge centralized databases on each and every person it can find,

    They already have the data. More than a few agencies track each of us. Why is this “centralized”? Have you ever heard of networks? What’s missing is some software, which is considerably cheaper than 10k+ millimeter wave imaging scanners.

    JA,
    Of course perhaps a better option would be mix the data with well trained dogs wandering about. We don’t have to really worry about box cutters any more. It’s just explosives and other chemicals, which dogs can screen. Boonton’s better scanner might be a not-yet-developed chemical sniffing device.

  26. Boonton says:

    Count the number of airports in your state. Then compare to the number of malls, stadia, and theaters. Count the number of exits and entrances for each. You think this is the same problem?

    Now you are just confusing the probability of an event happening with the difficulty of preventing the event. But this does lead to an interesting fact, you could demonstrate your profiling algorithm without having to worry about prevention. Simply set your profile loose and keep track of who it flags. If it flags all the people who later turn up in rampaging or terrorist events, then you may have something even if it also flags 20% of the population. Unfortunately you are going to have to wait a while as such events only come around now and then.

    At any cost. Riiight.

    Actually the cost of the current system is finite, and upgrading to scanners that go beyond simple metal detectors is finite as well. Plus it benefits from scale, once you have your scanner in place, there’s not much difference in cost between having everyone walk through it versus only a fraction of the people walk through it.

    Also I believe there is gradients between no-fly and fly with intenise search.

    They already have the data. More than a few agencies track each of us.

    Again I think you haven’t had much experiences with real life databases. A centralized database contining lots of information is not simply the sum of decentralized databases containing bits of information. When you have different databases you immediately start raising lots of serious problems in translating the data back and forth, even stupid things like listing people by their first and last name (i.e. Facebook) versus a unique number (tax id for the Feds, drivers license # for states say, maybe even birth date and mother’s maiden name for some financial institution) adds complexity. Centralizing would require stepping all over Federalism and wouldn’t even begin to address the fact that most of the data you think would be very important resides in private databases (Google, Amazon, Facebook, credit reports obscure blogs devoted to fringe topics etc.) The really interesting data is behind walls, such as your daily credit/debit card use…combining facial recognition software with the images caught by millions of security cameras that would enable one to build a megadatabase that would be able to create a picture of where you went almost every day of the year.

    But that’s not the current status quo, from the actual databases actually available today you couldn’t build a profile in the form you mention. There is no system that tracks whether or not you change churches or even take up gun ownership, for example. All you have to work with might be public internet posts you can gather from Facebook or Google (which by definition is incomplete since many people don’t bother much with Facebook), tax and social security data which is updated only quarterly and only captures employer and earnings, maybe you can buy credit reports but they too are only updated every now and then and do not contain a lot of detailed day by day data.

  27. Mark says:

    Boonton,

    The really interesting data is behind walls, such as your daily credit/debit card use…combining facial recognition software with the images caught by millions of security cameras that would enable one to build a megadatabase that would be able to create a picture of where you went almost every day of the year.

    Two things. Facial recog? Hello? That’s why an ID scan/entry at the point of search is interesting, you don’t need that tech. Second, you didn’t read the Orin Kerr link, Patriot act laws make some of that data available, such as credit records.

    And many of the other private databases you mention containing interesting records, google, Facebook, blogs, &c are well, publicly available even to feds.

    Also I believe there is gradients between no-fly and fly with intenise search.

    OK. How do you trigger gradiated search?

  28. Mark says:

    Boonton,

    Again I think you haven’t had much experiences with real life databases. […] When you have different databases you immediately start raising lots of serious problems in translating the data back and forth, even stupid things like listing people by their first and last name (i.e. Facebook) versus a unique number (tax id for the Feds, drivers license # for states say, maybe even birth date and mother’s maiden name for some financial institution) adds complexity.

    This you claim in a country in which the phone companies have given federal access to switch data and information for 50 years. Yet you believe, that VISA servers (for example) have not done the same.

    Now Mr Kerr thinks that the feds access to public (and not so public) data is a real privacy security concern on Constitutional grounds. You think the threat doesn’t exist. But … is that right? Do you really think that or is just a pose for rhetorical purposes?

  29. Boonton says:

    Two things. Facial recog? Hello? That’s why an ID scan/entry at the point of search is interesting, you don’t need that tech. Second, you didn’t read the Orin Kerr link, Patriot act laws make some of that data available, such as credit records.

    Consider what often happens after a famous crime. Footage comes out of various places the person was at before he was caught….a Wal-mart, crossing the street, etc. These individual databases of security footage are not centralling linked. The footage only becomes noted after the fact when various people realize person X whose in the news today happened to have been captured on camera before. If they were, though, you could build up pretty good ‘walking profiles’ of quite a few people which may produce some elements of a useful profile. For example, Person X may not appear on any jihadi social networks but he’s often seen having coffee with multiple different members at local Starbucks’s.

    OK. How do you trigger gradiated search?

    Again there are those who are on an absolute no-fly list as well as those who are triggered for searches, but they are allowed to fly provided nothing bad is found. TSA agents can also initiate a more detailed search based on a hunch and there’s an amount of random searches too. So right now the status quo is basically multiple tactics; universal screening, profiled screening, ‘on the ground hunch’ screening, and random screening (plus absolute no-fly lists which we don’t know much about so its hard to talk much about their merits or demerits).

    Anyway, let’s bring this to a close. Mark has all he needs to demonstrate profiling works. He simply needs to build up a list of people ‘flagged’ by whatever algorithm he thinks would work. If over the next few decades we collect the names of people who attempt terrorist acts we’ll check them against his profile list, if terrorists show up in his profile at rates significantly larger than would be dictated by pure random chance then he will be on his way to making his case. Since he has argued multiple times all the necessary info is publically available, he needs no special police powers to do this.

  30. Mark says:

    Boonton,
    That’s not true.

    Again there are those who are on an absolute no-fly list as well as those who are triggered for searches, but they are allowed to fly provided nothing bad is found.

    You have the former, but there is no checking for the latter. No ID/boarding pass is seen by the TSA search/screener.

    Mark has all he needs to demonstrate profiling works. He simply needs to build up a list of people ‘flagged’ by whatever algorithm he thinks would work.

    You mean if I go the work of the NSA/TSA I can demonstrate it works. Riiiight. In my copious free time.

  31. Boonton says:

    You have the former, but there is no checking for the latter. No ID/boarding pass is seen by the TSA search/screener.

    Fail: http://en.wikipedia.org/wiki/Secondary_Security_Screening_Selection

    You mean if I go the work of the NSA/TSA I can demonstrate it works. Riiiight. In my copious free time.

    Not at all, that’s the nice thing about an algorithm from publically available databases, you don’t actually have to do the work of screening. All you do is compile a database of individuals who would be flagged. Then you compare that to the general population (say it’s 5%). Then as actual terrorists or rampagers or nuts show up in the news, you note whether or not they appeared as flagged. Let’s say 50% of them do, since the odds of that happening by chance are nearly zero, you established your algorithm works.

    This would be the same procedure you’d follow if you claimed you had a system of, say, predicting rain or stock prices. The world doesn’t need to give you a farm or hedge fund to test your claim.

    I’ll even help get you started, David Brooks builds a mini-profile of many ‘rampage killers’ in the NYT (see http://www.nytimes.com/2012/07/24/opinion/brooks-more-treatment-programs.html?_r=1&partner=rssnyt&emc=rss). They all tend to be male, typically in their late teens to late twenties,

    Many of the killers had an exaggerated sense of their own significance, which, they felt, was not properly recognized by the rest of the world. Many suffered a grievous blow to their self-esteem — a lost job, a divorce or a school failure — and decided to strike back in some showy way.

    If you just selected males between 15 and 35 for screening you probably will do much better then pure random chance at picking up attackers. If you can reliably achieve 100%, then you have meet your goal of cost savings.

  32. Mark says:

    Boonton,
    So, is that your profile? 15-35? Or are you going to advance a more sophisticated one?

    I see, you want me spend my 45 seconds thinking about this. OK. I’ll get back to you.

  33. Boonton says:

    So, is that your profile? 15-35? Or are you going to advance a more sophisticated one?

    Let’s consider a purely dumb profile, flip a coin heads yes tails no. If you implemented this policy you’d screen about 50% of the population. Of terrorists, you’d expect 50% to be caught by this method and 50% to clear through it.

    So define that as a totally useless profile in that it does no better than pure chance in getting at your target.

    I would put forth to you my rather simple profile would be a dramatic increase. Make it all males. You’d still screen 50% or so but you’d probably catch 90% or more of terrorists.

    I’ll put forth that any ‘sophisticated’ profile you come up with that would screen less people than that could not match a 90%+ catch rate.

  34. Boonton says:

    Saw Holy Rollers this weekend on the tv. Good pic with that guy from the Social Network…does a good job playing a young Hasedic Jew who gets roped into an extasy smuggling ring and runs with it, based, they say, on a true story.

    It was, of course, a textbook example of the problem with profiling when your enemy is also trying to game your profile. Because no one would suspect a Hasedic Jew of smuggling drugs, they were almost never questioned, and even if they were the questioners could write off their nervousness as being due to the radical cultural differences.