So do you think the IAO would identify a group of people and watch whatever they do for patterns consistent with potential terrorist activity, or would they look at everything for terroristic patterns and then see what group is behaving that way?
It works a lot more effectively if you do the latter as opposed to the former.
Your concerns are very well placed, but they ignore the fundamental issue that any technology can be misused for good or evil purposes, and that if it can be done (which I am by no means sure will be the case), it will be done at some point. It's kind of hard to tuck the mushroom-shaped cloud back into the bomb casing and say "Whoopsie! Maybe we shouldn't've done that!"
It detects anomolous patterns that make utterly no sense on the surface level.
It is already used by catalog companies who on phone orders present you at the end with "specials" that the computer whips up just for you. Their acceptance rate on customers buying the spur of the moment "specials" has gone from under 10% to about 75% since they have begun using advanced algorithms which make seem to make no sense at all. That's just one example of how Mad Ave uses this technology to target specific consumers, and now this govt agency wants to take it to entire new levels in the war on terror.