A few points about closed circuit surveillance cameras, since their relative uselessness in the camera-festooned Times Square doesn't seem to have stemmed the call for yet more cameras as an anti-terror measure.
First, I think it's helpful to be clear just what we're talking about when we say "urban surveillance cameras." Lots of private businesses and apartment buildings have their own cameras trained at least in part on public spaces. And at this point, most of us are carrying around miniature cameras in our pockets 24/7 as well. I've read reports suggesting that the most promising video police obtained of the suspected bomber came not from the many CCTV cameras the city has in place there, but from a tourist who'd been taping in Times Square. These provide many of the same advantages as official surveillance networks—after a crime occurs, police can obtain and collate footage from the scene from the various owners—without creating a centrally controlled surveillance architecture. For the remainder of the post, I'll assume "cameras" means just such a citywide network of government controlled cameras, of the sort famously deployed in the U.K. and planned for New York—but it's worth noting that a city without these kinds of cameras is not necessarily a city without video evidence of crimes.
Second, while there will of course be the odd case one can find where cameras were instrumental in solving a crime, the research that's been done on public CCTV networks shows tha they're of stunningly little evidentiary or deterrent value. There are a few specific types of locations where the presence of cameras does seem to reduce, crime, or at least push it elsewhere. They seem to be fairly effective in parking lots. But on the whole, at the city level, they just don't work very well. In Britain, famously festooned with CCTV cameras, they're only rarely useful in apprehending street criminals, and the boroughs with more cameras don't seem to be any better at catching crooks than those with few. Anecdotal evidence can be beguiling here, because once you've created such a system of course the history of a few memorable apprehensions will involve the use of that system. If we gave cops lassos instead of guns and tasers, they'd end up lassoing a few crooks sooner or later too, but that hardly goes to show lassos are the right tool for the job.
Third, if citywide surveillance cameras are merely ineffective as a response to street crime, they're ludicrous as a response to the threat of terror. The point is, I think, well illustrated by New York Mayor Michael Bloomberg's invocation of the 7/7 bombings in London as an argument for installing an elaborate network of CCTV cams in New York: "You don't want to wait until 52 people are killed here and then say, 'Oh, now it's time to do it.' The trick is to learn by experiences, but it's other people's experience you'd like to learn by." What Bloomberg did not learn from the British experience, alas, is that 52 people were still killed. The billions spent on CCTV did nothing to deter the bombers, nor to disrupt their plan in action. The Times Square bomber, far from being deterred, chose one of the most recorded locations in the city as his target—and ultimately failed because of his own incompetence, not because of any of the dozens of cameras trained on the Square.
This kind of scenario, incidentally, presents the strongest case for surveillance cameras: A failed attack where you actually have a perpetrator to try and track down after the fact. London's cameras did indeed help out on that score after the second, failed attempt at a bomb attack on the transit system: Since they had intended to die with their victims, the terrorists hadn't bothered with countermeasures like disguises, something that might conceivably occur to a non-suicidal terrorist plotter in the future. Of course, those failed attackers were also seen by dozens of their intended victims, so there's little reason to think it would have been impossible to track them down but for the cameras.
Stipulating that the cameras did add some value in that rather unusual case, though, we need to step back and ask: Is this really the best security use we can make of a few hundred million dollars per year? An elaborate camera network that doesn't reduce crime, but might be of marginal benefit in tracking down perps after failed terror attacks by inept bombers? If we've gotten this disconnected from any rational cost/benefit analysis once the word "terrorism" is uttered, let's just start building enormous mousetraps made of gold and bait them with South Park DVDs; maybe we'll catch a few jihadis that way.
Finally, there's the question of privacy, which I leave for last because I don't actually think you can reject citywide camera networks on security grounds alone. Still, it's worth pushing back on the notion that there are no privacy concerns worth speaking of because, after all, the cameras are only trained on "public" places. Lying in the background of that argument is a rather crabbed notion of privacy that Daniel Solove has called the
"secrecy paradigm," and it assumes that privacy just means limiting the exposure of information that had otherwise been completely secret. But in practice, much of our privacy is not a function of the secrecy of information, but of its searchability and aggregability. There is a world of difference between knowing that any of your public behavior can be observed by others, and knowing that all of it is—that, indeed, a complete record of your public movements and actions can be automatically reconstructed from a central digital archive. Most of us probably don't mind shopping at "public" pharmacies full of indifferent strangers, but most of us would also be upset if a permanent record of our purchases were posted on the Internet with our names attached. And there's a difference, again, between merely being recorded and knowing that an automated behavioral analysis algorithm is apt to send up a red flag if any of your actions trigger a program's definition of "suspicious behavior."
To the extent that popular privacy discourse is saturated in the secrecy paradigm, it might be better to do away with "privacy" talk altogether, because the exercise of categorizing various kinds of information in a binary public/private schema may help while away the hours on a rainy Sunday, but it's not ultimately that interesting. The question we ought to be asking is whether and to what extent monitoring technologies facilitate social control. Sometimes that will be a price worth paying for security, but here, the case is quite weak.