5 ms·

I again want to highlight how you use the generic term "terrorism" as short-hand for a specific type of terrorism, specifically 'those people (in the middle eas

by dalke 10y ago

I again want to highlight how you use the generic term "terrorism" as short-hand for a specific type of terrorism, specifically 'those people (in the middle east and the arab world) have had, and a common thread to those is ethnicity/religion.' You haven't defined how to identify "those people". What ethnicity is common to Umar Farouk Abdulmutallab and Ziad Jarrah? So you must mean religion. Do you put Muhammad Ali and other Black Muslims from the US in the same category as "these people"? Do you distinguish "salafi/fundamentalists" from other Muslims? Some Muslims say that Ahmadiyya Islam isn't a Muslim faith. Should they be included? How easy do you think it would be for someone to hide their religious beliefs in order to bypass your screen? Nor have you identified the false positive rate. How many terrorists do you think there are, and how people many fit the screening category? If you define "terrorist" to exclude anything caused by a non-Muslim (anti-abortion murderers, anti-black murderers, Irish/Basque/whatever separatist) then that's like shooting bullets at a barn then putting the target around what you hit - there are no non-Muslim terrorists because you defined it that way. Now, I'm not saying you personally have to do this. I'm pointing out why profiling won't work. Because a quick run of the numbers shows that you will be overwhelmed by false positives. > Did I ever give the impression that any of these weights You gave the very strong impression that "ethnicity/religion" is on the top of the list. I've since pointed out two characteristics which are even more accurate, and you keep arguing that "ethnicity/religion" should be high on the list. Going back to your original question, which is "I don't see what's wrong with racial profiling." My answer again is that even setting aside the morality, the economics don't work out. What's wrong is you waste money for no gain and added frustration. > The very notion of "profiling" means you use all available info about likely attributes Here's where I know you have little experience with data mining. Every attribute you consider adds a new dimension to your search space. But there aren't that many terrorists, we have very little data in how to train the model, and most of that data is very noisy. Even in cases where we have lots of data, like schools, attempts to use models like VAM have produced what is at best confusing information, and at worst - and more often - is camouflage used to force certain political beliefs without real justification. How are you so sure that screening based on ethnicity/religion will make one whit of difference? How do you measure its effectiveness? When you say "from all publicly known cases" - how many cases do you think there are? 100? 1,000? How much statistical certainty do you think we can get from that? > That's (the "frustration of a TSA check" or similar minor event) is not historically (from all known cases) what such people have in their background. It's illegal for the TSA to profile, so there is no such "similar minor event" to use as a reference. Which cases were you thinking of when you said "all known cases"? What I mean are the more general examples of profiling, which have lead to anger, protests, riots, and killings. The Black Panther Party would not have started had black Americans not been routinely profiled. Or if you prefer, we use profiling to target our cruise missiles and drone attacks. Some of those kill wedding parties, babies, and other civilians. This makes some people angry enough to want to attack the US. But of course, this isn't a "minor event" like TSA screening. It's still an example of how profiling can induce new extremism. And again, how do you tell if a race/ ethnicity/ religion/ gender/ occupation/ nationality/ whatever screen is effective, when the base rate without such screening is zero?

coldtea 10y ago
>I again want to highlight how you use the generic term "terrorism" as short-hand for a specific type of terrorism, specifically 'those people (in the middle east and the arab world) have had, and a common thread to those is ethnicity/religion.' That's because this is what the "war on terror" predominantly targets and 9/11 the reason the whole modern TSA scheme was created for. >You haven't defined how to identify "those people". What ethnicity is common to Umar Farouk Abdulmutallab and Ziad Jarrah? Again, I don't understand the sudden naiveness. Why when one says "ethinicity" should be a factor, would anybody think that he means one and only ethinicity? Obviously it's a set of ethnicities, each with its own weight. Profiling is based on making a prediction heuristic with the data you have. If that data shows ethinicities X,Y,Z are predominantly involved in such attempts, you should include them in their profile. That (racial profiling) is not different than any other profiling factor. If profilers searching for a serial killer with a specific MO look predominantly at "males 20-50" that's not very different too. And males 20-50 are an even more broad category than ethnicity X,Y,Z (I'm saying this for people worried about "false positives" in racial profiling. Those are par for the course with ANY type of profiling, and besides the point). Also, noone is saying that profiling should be THE method of determining if one is what they look for. Incoming information, surveillance, following, interviews, checks etc will obviouslu also be used. >So you must mean religion. No, I mean ethinicity. But religion is an equally important factor. For example even an atheist with ethnicity X might want to avenge country Y if their countries have grudges. And conversely, someone with the same religion, but from a country that doesn't have any hostillity towards country Y, wouldn't be as likely to attack it. E.g. an Mongolian muslim compared to some Libyan muslim. >Do you put Muhammad Ali and other Black Muslims from the US in the same category as "these people"? Do you distinguish "salafi/fundamentalists" from other Muslims? Isn't the answer obvious? All those are factors that should be weighted in, too. >How easy do you think it would be for someone to hide their religious beliefs in order to bypass your screen? Even if it was easy (which would include living a "double life" for several years, being able to now have visible ties with said beliefs etc -- which some can just not do at all retrospectively), it would just mean that they particularly wouldn't be affected by said weight. Lots of others still would. >If you define "terrorist" to exclude anything caused by a non-Muslim (anti-abortion murderers, anti-black murderers, Irish/Basque/whatever separatist) then that's like shooting bullets at a barn then putting the target around what you hit - there are no non-Muslim terrorists because you defined it that way. I restricted the discussion to this, because it was a thread about racial profiling / TSA. Obviously the TSA is not involved much in looking for anti-abortion murderers or anti-black murderers. Nor I think they face (or should care) for Irish/Basque's seperatists. Ad obviously not every organization should care for all of those cases (e.g. Indian border control should not be much concerned with discovering Irish seperatists). That said, if one wanted to check for Irish seperatists, then ethnicity (being Irish) and religion (being, say, Catholic) should be good starting heuristics. >Now, I'm not saying you personally have to do this. I'm pointing out why profiling won't work. Because a quick run of the numbers shows that you will be overwhelmed by false positives. False positives are inherent in ANY profiling. That doesn't say anything against racial profiling in particular -- or even profiling in general. In fact, false positives are expected in profiling. The utility of profiling is that with it false positives, while numerous, are less than with just looking at the general population. And that false negatives are still good -- since most attackers are predictable in many "profiling" ways -- and have been historically so. So, yeah, you lose some outliers, but you are not guaranteed to get them without profiling either -- you will just make your job much more tedious. >You gave the very strong impression that "ethnicity/religion" is on the top of the list. I've since pointed out two characteristics which are even more accurate, and you keep arguing that "ethnicity/religion" should be high on the list. What, what would those characterists be that you've pointed out? I missed them. In any case, whether it's on the "top of the list" or in the top five or ten, they are still powerful factors to consider. >Going back to your original question, which is "I don't see what's wrong with racial profiling." My answer again is that even setting aside the morality, the economics don't work out. What's wrong is you waste money for no gain and added frustration. There's no moral question I can see. Not more than with any other type of profiling at least. It's just applying actual knowledge (historical priors) about such cases in your search. As for the more money, I'd say with profiling you spend LESS money (for the reasons I gave above: you've narrowed the search surface to what's MORE likely to be the case). >It's illegal for the TSA to profile, so there is no such "similar minor event" to use as a reference. Illegal doesn't mean much with regards to the actual practice. Despite being illegal they do it all the time, even subconsiously, but usually consciously too. >Which cases were you thinking of when you said "all known cases"? Known attacks. >Or if you prefer, we use profiling to target our cruise missiles and drone attacks. Some of those kill wedding parties, babies, and other civilians. This makes some people angry enough to want to attack the US. But of course, this isn't a "minor event" like TSA screening. It's still an example of how profiling can induce new extremism. I wouldn't say it's the profiling in those cases. It's the actual murdering of relatives/friends/copatriots (plus the whole without trial thing).

by dalke 10y ago

coldtea 10y ago

>I again want to highlight how you use the generic term "terrorism" as short-hand for a specific type of terrorism, specifically 'those people (in the middle east and the arab world) have had, and a common thread to those is ethnicity/religion.' That's because this is what the "war on terror" predominantly targets and 9/11 the reason the whole modern TSA scheme was created for. >You haven't defined how to identify "those people". What ethnicity is common to Umar Farouk Abdulmutallab and Ziad Jarrah? Again, I don't understand the sudden naiveness. Why when one says "ethinicity" should be a factor, would anybody think that he means one and only ethinicity? Obviously it's a set of ethnicities, each with its own weight. Profiling is based on making a prediction heuristic with the data you have. If that data shows ethinicities X,Y,Z are predominantly involved in such attempts, you should include them in their profile. That (racial profiling) is not different than any other profiling factor. If profilers searching for a serial killer with a specific MO look predominantly at "males 20-50" that's not very different too. And males 20-50 are an even more broad category than ethnicity X,Y,Z (I'm saying this for people worried about "false positives" in racial profiling. Those are par for the course with ANY type of profiling, and besides the point). Also, noone is saying that profiling should be THE method of determining if one is what they look for. Incoming information, surveillance, following, interviews, checks etc will obviouslu also be used. >So you must mean religion. No, I mean ethinicity. But religion is an equally important factor. For example even an atheist with ethnicity X might want to avenge country Y if their countries have grudges. And conversely, someone with the same religion, but from a country that doesn't have any hostillity towards country Y, wouldn't be as likely to attack it. E.g. an Mongolian muslim compared to some Libyan muslim. >Do you put Muhammad Ali and other Black Muslims from the US in the same category as "these people"? Do you distinguish "salafi/fundamentalists" from other Muslims? Isn't the answer obvious? All those are factors that should be weighted in, too. >How easy do you think it would be for someone to hide their religious beliefs in order to bypass your screen? Even if it was easy (which would include living a "double life" for several years, being able to now have visible ties with said beliefs etc -- which some can just not do at all retrospectively), it would just mean that they particularly wouldn't be affected by said weight. Lots of others still would. >If you define "terrorist" to exclude anything caused by a non-Muslim (anti-abortion murderers, anti-black murderers, Irish/Basque/whatever separatist) then that's like shooting bullets at a barn then putting the target around what you hit - there are no non-Muslim terrorists because you defined it that way. I restricted the discussion to this, because it was a thread about racial profiling / TSA. Obviously the TSA is not involved much in looking for anti-abortion murderers or anti-black murderers. Nor I think they face (or should care) for Irish/Basque's seperatists. Ad obviously not every organization should care for all of those cases (e.g. Indian border control should not be much concerned with discovering Irish seperatists). That said, if one wanted to check for Irish seperatists, then ethnicity (being Irish) and religion (being, say, Catholic) should be good starting heuristics. >Now, I'm not saying you personally have to do this. I'm pointing out why profiling won't work. Because a quick run of the numbers shows that you will be overwhelmed by false positives. False positives are inherent in ANY profiling. That doesn't say anything against racial profiling in particular -- or even profiling in general. In fact, false positives are expected in profiling. The utility of profiling is that with it false positives, while numerous, are less than with just looking at the general population. And that false negatives are still good -- since most attackers are predictable in many "profiling" ways -- and have been historically so. So, yeah, you lose some outliers, but you are not guaranteed to get them without profiling either -- you will just make your job much more tedious. >You gave the very strong impression that "ethnicity/religion" is on the top of the list. I've since pointed out two characteristics which are even more accurate, and you keep arguing that "ethnicity/religion" should be high on the list. What, what would those characterists be that you've pointed out? I missed them. In any case, whether it's on the "top of the list" or in the top five or ten, they are still powerful factors to consider. >Going back to your original question, which is "I don't see what's wrong with racial profiling." My answer again is that even setting aside the morality, the economics don't work out. What's wrong is you waste money for no gain and added frustration. There's no moral question I can see. Not more than with any other type of profiling at least. It's just applying actual knowledge (historical priors) about such cases in your search. As for the more money, I'd say with profiling you spend LESS money (for the reasons I gave above: you've narrowed the search surface to what's MORE likely to be the case). >It's illegal for the TSA to profile, so there is no such "similar minor event" to use as a reference. Illegal doesn't mean much with regards to the actual practice. Despite being illegal they do it all the time, even subconsiously, but usually consciously too. >Which cases were you thinking of when you said "all known cases"? Known attacks. >Or if you prefer, we use profiling to target our cruise missiles and drone attacks. Some of those kill wedding parties, babies, and other civilians. This makes some people angry enough to want to attack the US. But of course, this isn't a "minor event" like TSA screening. It's still an example of how profiling can induce new extremism. I wouldn't say it's the profiling in those cases. It's the actual murdering of relatives/friends/copatriots (plus the whole without trial thing).