The Net Delusion - Evgeny Morozov [58]
It’s not the implementation but the underlying principle that should have stood out. GreenDam is extremely invasive, taking a thorough look at the nature of the activities users engage in. It is programmed to study users’ computer behavior—from browsing websites to composing text files to viewing pictures—and try to prevent them from engaging in activities it doesn’t like (mostly by shutting down the corresponding applications, e.g., the Internet browser or word processor). For example, the color pink is GreenDam’s shorthand for pornography; if it detects too much pink in the photos being viewed, it shuts down the photo-viewing application (while photos of nude dark-skinned people, perversely, pass the civility test).
Most disturbingly, GreenDam also features an Internet back door through which software can communicate with its “headquarters” and share behavioral insights about the user under surveillance. This could teach other GreenDam computers on the network about new ways to identify unwanted content. GreenDam is a censorship system with immense potential for distributed self-learning: The moment it discovers that someone types “demokracy” instead of “democracy” to avoid detection, no other users will be able to take advantage of that loophole.
Think of this as the Global Brain of Censorship. Every second it can imbibe the insights that come from millions of users who are trying to subvert the system and put them to work almost immediately to make such subversions technically impossible. GreenDam is a poor implementation of an extremely powerful—and dangerous—concept.
Time to Unfriend
But governments do not need to wait until breakthroughs in artificial intelligence to make more accurate decisions about what it is they need to censor. One remarkable difference between the Internet and other media is that online information is hyperlinked. To a large extent, all those links act as nano-endorsements. If someone links to a particular page, that page is granted some importance. Google has managed to aggregate all these nano-endorsements—making the number of incoming links the key predictor of relevance for search results—and build a mighty business around it.
Hyperlinks also make it possible to infer the context in which particular bits of information appear online without having to know the meaning of those bits. If a dozen antigovernment blogs link to a PDF published on a blog that was previously unknown to the Internet police, the latter may assume that the document is worth blocking without ever reading it. The links—the “nano-endorsements” from antigovernment bloggers—speak for themselves. The PDF is simply guilty by association. Thanks to Twitter, Facebook, and other social media, such associations are getting much easier for the secret police to trace.
If authoritarian governments master the art of aggregating the most popular links that their opponents share on Twitter, Facebook, and other social media sites, they can create a very elegant, sophisticated, and, most disturbingly, accurate solution to their censorship needs. Even though the absolute amount of information—or the number of links, for that matter—may be growing, it does not follow that there will be less “censorship” in the world. It would simply become more fine-tuned. If anything, there might be less one-size-fits-all “wasteful” censorship, but this is hardly a cause for celebration.
The belief that the Internet is too big to censor is dangerously naïve. As the Web becomes even more social, nothing prevents governments—or any other interested