Please note that by playing this clip YouTube and Google will place a long term cookie on your computer.
When the CIA's Chief Technology Officer, Gus Hunt, spoke at a GigaOm conference last week in New York, he told the audience that the CIA collects and stores every bit of information it can get its hands on, and then deploys data mining tools to extract meaning from the swirling and ever growing matrix. (You can see the slides from his talk here.)
"The value of any piece of information is only known when you can connect it with something else that arrives at a future point in time. Since you can't connect dots you don't have, it drives us into a mode of, we fundamentally try to collect everything and hang onto it forever," he said.
Hunt's talk, embedded above, works from the assumption that data mining can identify terrorist threats. But that's a faulty, if widely held, view. The National Academy of Sciences studied the question back in 2008 and found that it is "not feasible" to successfully deploy data mining in terrorism prevention. On the other hand, the aggregation and mining of data about all of us is all but guaranteed to have a substantial impact on our privacy and civil liberties.
Data mining for terrorism prevention doesn't work because, thankfully, there aren't very many terrorists attacks. The terrorists attacks that do occur are so wildly different, and are executed by people with disparate backgrounds and motivations, using different tools, who organize and plot and execute the attacks with varying styles. So it's basically impossible for computer scientists to teach computers how to look for information that may identify terrorists: there is no single 'terrorism pattern' to teach the machines to watch out for.
The risk of 'false positives' is also huge, because there are high stakes when the government incorrectly identifies someone as a terrorist -- ranging from getting placed on a no fly list to being detained or killed. The CIA's own 'signature strikes' program uses behavioral profiling to target for death people the agency can't even identify, but simply assumes are terrorists based on a set of data points.
Here at home, DHS is deploying data mining to control air travel, using a system it calls the Automated Targeting System. The National Counterterrorism Center collects information about every US citizen and mines it for so-called 'terrorism' indicators, and also feeds possible kill targets to other government agencies.
Does data mining for terrorism work? Hardly. Will that stop the government from doing it? Absolutely not.
But it's not all failure: the government's obsession with data collection and mining is very effective at enforcing social control.