A more high-tech instance of this problem is encrypted communications. If I'm worried that The Man will discover that lemuru and I are plotting to smash the state, we can exchange keys and communicate using strong encryption. We can make it practically impossible for anyone to find out what our emails say for the next several years. But it's substantially more difficult to disguise where a message originates, or where it's going. If carnivore (or a good old wiretap) uncovers the fact that we're taking some effort to make our email unreadable, that in itself can indicate that there's a conspiracy afoot. They can then begin investigating our activities through other means.

Edibleplastic touched on one solution to this problem earlier: if everyone had a tarp in their car, then the tarp would not be a useful source of information. The same would happen if everyone used strong encryption on their email. The spooks would have no idea whose mail to overclock their supercomputers cracking. Rather than trying to neutralize the persistent signal, we immerse it in undifferentiable noise.

This is the primary reason I use PGP.

Now, there's a substantial barrier to getting everyone to put a tarp in the back of their car. If we assume that there's a constant, unchanging number of thefts, then if everyone has a tarp the chances of having your car broken into (a major hassle, even if nothing is stolen) are completely random -- the number of thefts divided by the number of cars. But if only a few people have tarps, then those cars will get broken into, and the ones without will be left alone. It's like those no radio signs.

In this example, one flaw is that the number of thefts is not necessarily constant. If every car has a tarp, then theivery might become unrewarding enough to vanish completely. Then everyone benefits. But if there are even a few thefts, then one stands to benefit even more from having everyone else use a tarp, and leaving one's own car without one. Of course, if everyone thinks that way there will be no tarps. This is an illustration of a large-scale prisoner's dilemma, in which a few people who defect can cause widespread breakdown. A less abstract formulation is the free rider problem.

In the case of encryption, though, the situation is different. It's equally hard to break into a car with a tarp, but it's a heck of a lot harder to spy on encrypted email. And while those who encrypt run some risk of getting fingered by the NSA, they eliminate a much greater one that some 12 year-old reprobate with a packet sniffer will use the chance to giggle over their love letters. So, everyone who adopts encryption both gains an immediate, personal benefit and contributes to a decrease in the associated risks. This is also helped by the fact that internet users feel some altruism towards their community; the recent Jam Echelon Day, directed against the international espionage infrastructure ECHELON, was only the latest in a long line of attempts to overload keyword-based spying in a way that offers no personal benefits to the participants.

Communities can sometimes overcome the persistence of knowledge problem. The way to do it is to find a system that benefits individuals immediately, as well as helping out the group once it reaches critical mass.

(none of this really helps storm. her only option would have been to anticipate this contingency far in advance, and cause New York to fog over irregularly for a long period of time.)


note: Thanks to caseyhb for pointing out a lack of clarity in my discussion of similarities between defecting prisoners and free riders. It has been amended. I hope