CMU study analyzes what China deletes
Share with others:
Iodized salt isn't typically considered a dangerous chemical compound, but to Chinese social media censors hoping to squash rumors that the substance could prevent radiation poisoning following last year's nuclear disaster in Japan, the term became downright unspeakable.
"The Chinese government came in, put their foot down and said don't believe these rumors. After that, iodized salt became a sensitive topic and it was highly likely a message would be deleted if it discussed salt," said David Bamman, a Ph.D. candidate at Carnegie Mellon University's Language Technology Institute and co-author of the study, "Censorship and Deletion Practices in Chinese Social Media."
"Iodized salt," "brainwash" and "Falun Gong," a religion declared a cult and banned by the Chinese government in 1999, were among thousands of terms that triggered censorship when used in social media last year, according to the CMU study.
Mr. Bamman, along with Brendan O'Connor, a Ph.D. student in CMU's Machine Learning Department, and CMU Professor Noah Smith focused on content from Chinese micro-blogging service Sina Weibo to gauge the likelihood that messages with politically charged words, names and phrases will be deleted by either private or public censors.
Sina Weibo is one of the most popular microblogging sites in China and one of the most active in the world, surpassing Twitter's usership with more than 200 million accounts. In 2009, the Chinese government blocked citizens' access to Twitter and Facebook.
"A lot of studies have focused on censorship that blocks access to Internet sites, but the practice of deleting individual messages is not yet well understood," Mr. Smith said. "The rise of domestic Chinese micro-blogging sites has provided a unique opportunity to systematically study content censorship in detail."
The team studied approximately 57 million messages sent on Sina Weibo between June 27 and Sept. 30 last year, using a programming application the company provides to developers to create related content.
Three months later, through the same application, the researchers checked a random set of "Weibos," as well as a set that included politically charged messages, to see if any had been deleted.
When the reviews were complete, they found that 212,583 out of more than 1.3 million checked messages had been deleted, or more than 16 percent.
They also found 54 percent of checked messages sent from Tibet had been deleted and messages featuring "sensitive" political terms were highly likely to be deleted.
Although the political sensitivity of many terms, such as iodized salt, depended on a particular situation, the team was able to narrow down a list of 295 terms and names with a high probability of being censored. About 33,363 total messages were found to include at least one sensitive term and 5,811 of those, 17.4 percent, had been deleted.
The study was published in the March issue of the peer-reviewed online journal First Monday.
Considering other blocks China has imposed on Internet content through the Golden Shield Project, an Internet surveillance and censorship system also known as the Great Firewall of China, Mr. Bamman said the government has succeeded in its overall mission of preventing citizens from using the Internet to sway political discourse.
"What they hope to do is to try to prevent massive organizations of people and problems they saw in the Middle East over the years through social media. By preventing people from discussing certain things in China, they have a way to effectively suppress that," he said.
However, others argue that China's online community is hungry for more information and looking for ways to beat the system, despite being part of a society that has industrialized censorship.
Paris-based social media research company Synthesio's 2011 report, "Social Media and Censorship in China," quotes an anonymous Chinese tech veteran and blogger who says the government employs hordes of citizens to moderate sites for content that could fit criteria for the chopping block.
Dubbed the "50 Cent Army" or "50 Cent Party" for the price of each deleted post, the blogger said workers use automated keyword filtering systems as well as their own judgment when monitoring message boards, specific users or when seeking out posts on specific topics.
The blogger added that every content-related Internet service provider in the country is required to hire full-time employees to monitor content and employees must report to the government's Internet Propaganda department. Sina Weibo CEO Charles Chao said that company employs 100 censors, which the CMU study called a "low estimate."
Despite the constant surveillance, Chinese citizens aren't afraid to go out on a limb online, said Ben Farkas, Synthesio's director of U.S. sales and operations. He pointed out that during a breach in China's firewall that allowed Chinese citizens to access Google's Google+ social media service last week, hundreds of messages were sent to President Barack Obama's page decrying censorship and asking him to support the cause.
"It's kind of cool to see they took advantage of that time to say they're different from their government and are being suppressed by the wall," he said.
Even with the blocks, Chinese citizens are finding ways to get around censors to send political messages, said David Wertime, co-editor of the e-magazine Tea Leaf Nation, which focuses on social media and censorship in China.
He said citizens regularly use slang words, homonyms and Romanized letters to refer to potentially sensitive topics and names. More determined citizens also attach digital images to Weibo messages that include text blocks not searchable by censors.
And while Mr. Wertime seemed concerned that the Chinese government's plans to require real names on all Sina Weibo accounts by March 16 could discourage open discussion, he said he doubted the most vocal members of the Sina Weibo will back down when the day comes.
"People who aren't familiar with Chinese social media would be shocked at the level of candor and how often people stick their neck out to give frank and open comments, even if they're counter to the party line," he said.
Correction: (Published March 9, 2010) An earlier version of this story misidentified one way Chinese citizens use symbols and codes to outfox censors.
First Published March 8, 2012 12:00 am