[WS:] I agree with the general drift of your argument, but that is not how intelligence interception works, I am afraid. The procedure can be described as casting widest possible net and then successive filtering out of the noise. That is, they record all enemy communication they can and then have an algorithm looking for certain key words. If the algorithm finds them , that piece of communication if selected for a review by a human reader whose role is to confirm that the topic of that communication may be of interest. If yes, then the communication is forwarded to higher level analysts. This at least was the process when I was marginally involved in it (I was training the low level readers) and I am pretty sure that there has been substantial progress in voice recognition software which allows greater automation. But I think it wrong to assume that this a dumb algorithm based process that decides fates of unsuspecting victims.
Those guys want the real thing, not noise that looks like the real thing.
Of course, mistakes are made but I would bet that they err on the side of underidentification than overidentification. That is, they are more likely to miss something of interest than take garbage as something of interest.
The reason for it is that it is a Library of Babel out there http://www.thecriticalpoint.net/index_files/libraryofbabel.pdf. Most of what they are getting is noise and they simply do not have the resources to check everything that remotely looks suspicious. The way the FBI handled the Tsaranev case (Boston bomber) is speaking volumes. They received positive intelligence about this guy and checked him out, but did not find anything derogatory so they let him go. Big mistake! One may ask why did not they pursue the lead further, and I think the answer is rather simple - they do not have human resources to pursue every lead, so if they see a low probability of success they just drop the ball.
You also need to take into account that government information gathering in this country (not just spying but collection of statistics as well) is rather decentralized vis a vis most other countries, This means that multiple agencies are responsible for collecting different types of information, and they are not eager to share that information with other agencies for a variety of reasons. It does not means that sharing does not take place, but the process is not automatic and faces considerable institutional obstacles. The image of omniscient spy agencies is a myth fabricated by spy movies and conspiracy theorists.
I just wish that people were more thoughtful about the nature of intelligence gathering - which is LEGITIMATE and NECESSARY given the world we live in - and involves high level of uncertainty about defining goals, high level of information asymmetry (i.e. one side knowing more than the other), high stakes in getting real results, and a great deal of balancing acts they must do avoid ruffling feathers of a great number of stakeholders, from foreign governments, to local power players, to legal constraints, and to public opinion makers. This is, of course, not to say that there is a potential danger of this process getting out of hand, but spinning fantastic stories about the supposed capacities of that process will do little to prevent that danger.
PS. Based on what I read in the news, the process did have judiciary and congressional oversight so it was not spooks running amok.
-- Wojtek
"An anarchist is a neoliberal without money."