With over 300 million users sending 100 million messages a day, or 70,000 per minute, Sina (NASDAQ:SINA) Weibo, the "Twitter of China," is one channel filled with government criticisms. To combat it, China and Sina have censors in place -- but are they fast enough?
That's what five professors sought to understand in a recent study.
Dan Wallach (Rice University) and company primarily studied 3,500 users whose posts were deleted more often than others -- because they assumed they'd be more likely to be censored in the future. Over the course of 15 days in 2012, these users experienced about 4,500 deletions per day, or 12% of all posts. Digging deeper, the results are more revealing.
In their data set, they observed 5% of all deletions occurring within eight minutes of posting, and 30 percent within 30 minutes. Overall, 90% of deletions happened within the day. In other cases, post deletion happened several days later. Based on these results, Wallach and company point to several hypotheses.
First, as most deletions happen within 5-10 minutes of posting, Weibo must be censored near real-time. And if an average censor can scan around 50 posts a minute, that calculates to 1400 censors to handle the 70,000 posts each minute. If each censor works 8 hours a day, that means there's a total of 4200 censors.
Of course, given the speed of some deletions, Wallach and company suggest that Weibo must have a number of techniques in place.
The professors also came to several other conclusions:
- Censorship dipped slightly at 7pm, when the national evening news airs on television.
- Commonly deleted phrases contained phrases such as "support Syrian rebels," "Lying of gov. (Jixiang)," "One-Child policy abuse," and "group sex."
- Mass removals were fastest across posts combining popular events and sensitive themes, like "police" and "sex scandal."
- Frequently censored users are typically censored the fastest.
- Observing a 24-hour cycle, censors are less active at night. Presumably, fewer censors are working at that time. There's also a backlog each morning. Censors seem to catch up by late morning or early afternoon.
Given the size of authorities, Wallach and company wondered why all posts weren't blocked for 30 minutes -- which should be enough time for the "formidable censorship machine" to properly delete sensitive posts.
Moving forward, the professors say they have several other questions to study. One on their mind: How does Weibo prioritize content for deletion?