Why it’s so hard to tell porn spam from Chinese state bots

But this spammy content may have nothing to do with the Chinese government after all.,according toa report released Monday by the Stanford Internet Observatory. “While the spam drowned out legitimate content related to the protests, there is no evidence that it was designed to do so, or that it was a deliberate effort by the Chinese government,” wrote David Thiel, l author of the report.

Instead, they were probably just the usual commercial spam bots that have plagued Twitter forever. These particular accounts exist to attract the attention of Chinese users who go to foreign networks to access porn.

Therefore the “significant increase” in the spam was just a coincidence? The short answer is: very likely. There are two main reasons why Thiel doesn’t believe the bots are linked to the Chinese government.

First of all, these accounts have been posting spam for a long time. And they sent even more tweets, and more regularly, before the protests erupted, according to an analysis of activity data from more than 600,000 accounts from Nov. 15-29. Another analysis shows they also continued to spam even as talk of the protests died down.

Check out these two charts (for reference, the protests peaked around Nov. 27):

So is it fair to feel like spam activity increases during protests? This chart shows that many more bot accounts were actually created in November:

But Thiel points out that content moderation takes time. People tend to ignore the effect called “survival bias”: Older spammy content and accounts are constantly being removed from the platform, but researchers don’t have data on suspended accounts. So a chart like this only shows accounts that Survived Twitter spam filters. That’s why the November spike seems so important: these are new accounts most recently created to replace their deceased peers and are still standing, but not all of them will survive, so they wouldn’t be there if we were to revisit this graph in, say, a few months. In other words, if you were to analyze the data right after the protests, it would certainly be seem thisthis type of spam just started recently. But that’s not necessarily the whole truth.

Secondly, if the spam accounts were meant to bury information about the protests, they did a pretty poor job. While the escort ad spam featured many Chinese city names as keywords and hashtags, Thiel found that they did not target the hashtags actually used to discuss the protests, like #A4Revolution or #ChinaProtest2022,” which would you suppose the government would be interested in jumping on it if they tried to shut things down,” he tells me. there’s no spam to speak of in there”.