Lost and Found: Collections of Censored Posts on Sina Weibo

//Ran Liu, a graduate student in Sociology at the University of Pennsylvania, discusses China’s censorship of Weibo through an analysis of NGO ProPublica’s new “China’s Memory Hole” project.

Most netizens in China have complained about the “Little Secretary,” the censorship mechanism on the most popular Chinese microblog service Sina Weibo. Content may disappear at any minute only to be replaced by an ambiguous notice that a “post is inappropriate to display publicly.”  Users find it difficult to anticipate whether their posts will live the day as censorship logic is, while sometimes predictable, more often than not quite elusive.

Scholars are exploring the mechanisms of censorship on the Chinese Internet to better discern Chinese censorship practices. For instance, a recent study at Harvard University led by Gary King discovered that messages discussing or invoking collective actions are more likely to get censored. That is to say, “censorship is oriented towards attempting to forestall collective activities that are occurring now or may occur in the future.”[i] A new book from China Specialist Jason Ng also discusses key words blocked on Weibo.

The new project “China’s Memory Hole,” by the New York based NGO ProPublica, provides a more intuitive and interactive demonstration of China’s Internet censorship. From July 24to August 4, 2013, ProPublica tracked 100 Weibo accounts to detect censored images, collecting a total of 7,972 posts. Of the posts sampled, 527 were deleted by the Little Secretary. Although ProPublica did not randomly select the tracked Weibo accounts,[ii] indicating that their results should not be generalized to the larger population, we can still draw some interesting observations by looking closer at the project’s censored photo database.

One such observation applies to ProPublica’s categorization of the censored images, which could be refined. The project divides the censored images into ten categories: Political Speech, Long Text, Dissidents, former Chinese politician Bo Xilai, Public Figures, Protests, celebrity faith healer Wang Lin, Scandals & Corruption, Censorship, and Cartoons & Humor. These categories cause some issue as they sometimes overlap with each other, and they do not clarify the possibly diverse reasons for the images’ censorship. For example, the 156 images in the largest category, “Political Speech,” cover a wide range of topics. Posts include historic photos of the Korean War, a portrait of Stalin, a sarcastic comparison of President Obama and Chinese officials, a photo of Zeng Chengjie’s daughter on hunger strike, pictures of forced demolition, and screenshots of external news reports on domestic issues.

An improved classification method could be based on the targets and objectives of the posts. For example: Who or what are the posts criticizing? What are, or might be, the intentions of the posts? If left uncensored, what influences may the images have? These broad categories could then be further broken down. The targets of the posts could be divided into different levels, including the Chinese communist party, the legitimacy of the one-party authoritarian regime, the central political leaders, the local government officials, particular incidents violating human rights, or controversial state policies. Intentions could be classified as unintentional posts on non-political issues, intentional political criticism, and intentional calls for collective action.  This preliminary two-dimensional framework would contribute to an easier understanding of the logic of censorship.

Under this new framework, we can reclassify most of the censored images to identify the real reason leading to their censorship. For example, in ProPublica’s study, the 55 deleted images of Bo Xilai and 41 deleted images of Wang Lin were not related to calls for collective actions, but were removed primarily because most images were group photos featuring former high officials. More posts showing no signs of invoking collective action were also censored. Such posts included harsh critiques on the famous entrepreneur Li Kaifu, a news report on the rebellious son of General Li Shuangjiang, a report from Hong Kong-based Ta Kung Pao on Chinese high-level official inviting businessmen to dinner and offering them human breast milk, and stories and caricatures about the Korean War against the official propaganda. Some posts were deleted because they proved to be rumors, however, it still worth noting that rumors regarding non-political issues such as genetically modified food are freely spread without censorship, therefore the target of the rumor might be crucial.

Despite some issues with the project’s categorization, ProPublica’s “censorship of censorship” category is quite interesting. It appears the censors are particularly sensitive about discussions and posts opposing censorship. The 28 censored images protesting censorship include screenshots of already deleted posts, and notifications of posts being deleted. In some instances online censorship proved to be more conservative than that of traditional newspapers. Images of a local newspaper report  about a city’s possible lift on the Facebook and Twitter ban, and an article titled “Myanmar Lifts Ban on Facebook, which Only Remains Closed in Four Countries” were both censored despite their publication in Chinese newspapers. These cases may result from a concentration of censorship resources on new media. Due to this, content published in traditional media, especially local newspapers with limited influence, can survive censorship if it does not draw too much attention. Further research on this topic may unveil more possible differences between the censorship mechanisms and resources in new media versus traditional media.

This project has provided scholars a lively exhibition of censored images and retrieved valuable data for further analysis. In addition to ProPublica’s censored image database, Hong Kong University (HKU) has established an archive called Weiboscope that automatically detects deleted posts and records the exact time it became censored. Recent research based on HKU’s database discusses homophones and puns created by microbloggers to circumvent the censorship. This collection provides a robust sample for future research, as it is complete and less biased than ProPublica’s image database. As projects continue to generate collections of censored microblogging posts, scholars will be able to better discern some of the mechanisms behind Chinese online censorship.

//Ran Liu is a graduate student in Sociology at the University of Pennsylvania. Her research areas include social movements, non-governmental organizations and the influence of information technology on state power and civil society.


[i] King, Gary, Jennifer Pan, and Margaret E Roberts. 2013. How Censorship in China Allows Government Criticism but Silences Collective Expression. American Political Science Review 107, no. 2 (May): 1-18. Copy at http://j.mp/LdVXqN
[ii] The Weibo accounts tracked were selected from a subset of users who previously posted content that was later censored. The 100 users tracked have a minimum of 2,000 followers, and identified themselves as journalists and lawyers.

Featured Photo Credit:AttributionSome rights reserved by IsaacMao

Leave a Reply