Some Larger Questions Behind the Bush Abuses
About three years ago, while my team was working on an e-mail system for a large financial corporation, we met with the makers of a product that made responding to customer e-mail more efficient by applying fancy logic to automatically match the customer's e-mail to our set of pre-defined answers. The package applied Natural Language Programming (NLP) and Artificial Intelligence (AI) logic to the e-mail text, and would continually learn, as human beings -- customers and call center agents -- confirmed or disconfirmed its guesses. The result was that the application could find relevant information where humans would not have guessed -- e.g. sometimes as little as the tense or conjugations used could be relevant factors in understanding what the customer wanted to know.
During this particular meeting, the CEO and founder of the company, an ex-Israeli Intelligence agent, dropped the fact that their product was used by one nation's government to monitor large amounts of e-mail, looking for clues to anti-government activities. Of course he would not tell us which nation, though my friend Sudhir, who is much better at reading between the lines than I, later told me with some confidence that it was the Phillipines. In any case, it went without saying that it must be a relatively totalitarian regime -- one without anything like the civil rights protections and concern for individual liberties of the United States.
Since then we have learned that similar systems are being used to monitor thousands of communications of Americans with persons overseas without a warrant. The Bush administration has tried to lead citizens to believe that these are only calls involving al Qaeda suspects, but we can logically conclude that the surveillance has extended well beyond the kind of linkages we would normally associate with such phrases, let alone the legal definition of "probable cause." Further, this sort of data minining is aimed less at monitoring known suspects and more at *identifying* suspects -- from data trends human beings alone would normally not notice. Putting aside, for the moment, the apparent lack of productiveness of this NSA tapping and the administration's self-serving deceptions, there *should* be a robust debate about this topic, involving the administration, the Congress, and ultimately, the American people.
That is, given that we can have in place checks and balances consistent with the Constitution and a reasonable barrier to abuse, how willing should we be to let software cull vast numbers of communications for patterns -- *before* the communicating parties are likely suspects in any criminal behaviors -- in order to diminish the odds of egregious crimes? What kind of "hit rate" would be needed to justify the intrusion? If no human looked at particular message (and the old "corpus" of messages would eventually be destroyed) until computer analysis revealed that at least party was 40% likely to be involved in planning terrorist activities, would it be acceptable for the computers to monitor *all* voice and email communications? 60% likely? 90% likely? If the system could identify an estimated 2 terrorist planners a year, would that justify computers sifting through everyone's electronic communications? How should we modify the laws (or even the Constitution?) to deal with these scenarios?
One thing that I hope we all agree upon is that we should not simply give the President the ability to describe some scenario as being "at war" and subsequently ignoring any laws or checks as long as he simply tells us it's in the interest of national security. But we should be making some open decisions based on a slippery continua of rights and safety. Where would you draw the line?
During this particular meeting, the CEO and founder of the company, an ex-Israeli Intelligence agent, dropped the fact that their product was used by one nation's government to monitor large amounts of e-mail, looking for clues to anti-government activities. Of course he would not tell us which nation, though my friend Sudhir, who is much better at reading between the lines than I, later told me with some confidence that it was the Phillipines. In any case, it went without saying that it must be a relatively totalitarian regime -- one without anything like the civil rights protections and concern for individual liberties of the United States.
Since then we have learned that similar systems are being used to monitor thousands of communications of Americans with persons overseas without a warrant. The Bush administration has tried to lead citizens to believe that these are only calls involving al Qaeda suspects, but we can logically conclude that the surveillance has extended well beyond the kind of linkages we would normally associate with such phrases, let alone the legal definition of "probable cause." Further, this sort of data minining is aimed less at monitoring known suspects and more at *identifying* suspects -- from data trends human beings alone would normally not notice. Putting aside, for the moment, the apparent lack of productiveness of this NSA tapping and the administration's self-serving deceptions, there *should* be a robust debate about this topic, involving the administration, the Congress, and ultimately, the American people.
That is, given that we can have in place checks and balances consistent with the Constitution and a reasonable barrier to abuse, how willing should we be to let software cull vast numbers of communications for patterns -- *before* the communicating parties are likely suspects in any criminal behaviors -- in order to diminish the odds of egregious crimes? What kind of "hit rate" would be needed to justify the intrusion? If no human looked at particular message (and the old "corpus" of messages would eventually be destroyed) until computer analysis revealed that at least party was 40% likely to be involved in planning terrorist activities, would it be acceptable for the computers to monitor *all* voice and email communications? 60% likely? 90% likely? If the system could identify an estimated 2 terrorist planners a year, would that justify computers sifting through everyone's electronic communications? How should we modify the laws (or even the Constitution?) to deal with these scenarios?
One thing that I hope we all agree upon is that we should not simply give the President the ability to describe some scenario as being "at war" and subsequently ignoring any laws or checks as long as he simply tells us it's in the interest of national security. But we should be making some open decisions based on a slippery continua of rights and safety. Where would you draw the line?
posted by peterga at 11:31 AM
0 comments
