Study: Chinese Censors Police Social Media with Blinding Speed
While Chinese hackers were topping global headlines for poking around the computer systems of the United States' most valuable tech and media companies -- and movie studios -- a team of American researchers was busy carrying out its own clandestine inquiry into Chinese tech territory.
A new study published on Monday by independent researcher, Tao Zhu, and four U.S. academics details the workings of China’s most politically sensitive online activity: social media censorship.
Analyzing 2.38 million posts on China’s most popular social media platform, Sina Weibo -- a service similar to Twitter -- the researches break down the methods used by censors to police the millions of comments made by users on the site every day.
China is known for having the world’s most comprehensive and sophisticated Internet censorship system. But what’s often left out of the conversation is that it’s rarely the government itself doing the censoring when it comes to social media. Rather, the companies offering the services -- Sina, Tencent, NetEase, all of which offer their own ‘Weibo’ or micro-blog platform -- operate aggressive operations of self-censorship as part of a tacit agreement with the government, which essentially says: “we’ll let you stay in business, if you make sure your users don’t cause us too much trouble.” Government censors are presumed to only get involved when they feel the service providers aren’t doing a good enough job of scrubbing their portals clean of offending social and political commentary.
The main takeaway from the inquiry is a palpable sense of awe over how fast censors are able to filter the service given the massive volume of postings. With Weibo messages posted at an average rate of 70,000 per minute, the researchers found that “nearly 30 percent of the total deletion events occur within 5-30 minutes, and nearly 90 percent of the deletions happen within the first 24 hours.”
It goes on: “Considering the big data set that Weibo has to process, the speed -- especially at the 5 to 10 minutes peak -- is fast, especially [since] it cannot be processed in a fully automated way.”
With this speed in mind, the researchers speculate that censors likely use automated lists of sensitive keywords and prioritize users who have a history of politically affronting activity. When they discover an offending post, they delete it and quickly chase down all of the retweets -- usually eliminating them all within five minutes.
Only 10 percent of the posts aren’t deleted until after 24 hours. Those posts typically contain clever neologisms invented by users to sneak past the censors’ automated keyword filtering. For example, last year, many micro-bloggers employed the word “tomato” to refer to disgraced Chinese leader Bo Xilai, because of the similarity of the pronunciation of one element of Chongqing, the city Bo ruled, and the Chinese word for the color red. “Bo” also happens to mean “thin” in Chinese, so some users briefly used the expression “the not thick king” to make reference to the fallen political star.
The researchers say that once Chinese censors discover the widespread use of a new code word, they quickly add it to their automated list and scan and delete all past instances.
To prevent users from finding sensitive information, they also constantly update a list of key words that are rendered unsearchable on the service.
When a user is highlighted as a repeat offender, their entire account is deleted. The researchers say this happened to 300 of the 3,500 accounts they tracked for the study.
Pity the Chinese tomato farmer attempting his first foray into social media.