Thread: Search for Beijing/Shanghai/other cities in Chinese on Twitter and you'll mostly see ads for escorts/porn/gambling, drowning out legitimate search results.
Data analysis in this thread suggests that there has been a *significant* uptick in these spam tweets.
I searched for 北京 (Beijing) today (11/28 ~5am Beijing time) and identified accounts with tweets that show up in the "Latest" search results.
Vast majority (>95%) of these are spam accounts. They tweet at a high, steady rate throughout the day, suggesting automation.
Then I looked at the number of tweets by each account over time.
Interestingly, more than 70% of these spam accounts only started tweeting like crazy recently.
The rest (~top 90 in this plot) seem to have been spamming consistently for a while.
This is also true when I search for 上海 (Shanghai).
Note that these are just a small sample of the search results -- go and search 北京/上海 and you'll see new spam tweets come up every few seconds. So the number of spam accounts is way larger than a few hundred.
A technical point: the explosion of spam is not an artifact from sampling search results today.
When I look at the tweeting pattern of spam accounts that have tweeted earlier this week (instead of today), there is no steep increase. They likely reflect basal level of spamming.
This has been a quick and dirty analysis, feedback/questions highly welcome!
Data source —
Twitter search results: web search, “Latest” panel
# of tweets: Twitter API Tweet Counts
A thread on TikTok-Douyin decoupling (or lack thereof) thru some fun OSINT sleuthing:
Having read recent reports on ties between TikTok and DouYin(/Bytedance), which mostly relied on indirect evidence, I wanted to see if I could find some more direct data. (1/n)
Last year, I stumbled upon a group of TikTok accounts used by China-based Bytedance software/testing engineers to test various aspects of the app. This week, I expanded the list to ~1,300 by looking at accounts followed by/following known accounts. (2/n)
The accounts post public videos of their workstation/laptop/office space with various TikTok functionalities (Q&As, stickers, effects, etc).
The videos sometimes capture Bytedance IDs, office spaces w/ logos, and most interestingly, computer screens w/ codes and messages. (3/n)
Back to this analysis, I’ll admit it is quick and dirty: I got the idea at 2AM and stayed up till 4 to make the graph. Of course there are caveats, many of them I addressed in the original tweet.
But some other caveats/criticism:
1. Correlation is not causation.
Great point, and definitely something I thought about. It is difficult to properly show causation using complex, real-world data like this. Definitely something to try. But in the meantime...
...this correlation, plus 1) previous studies showing how Trump’s rhetoric emboldened some to engage in racist behavior, 2) recent anti-Asian discrimination in the real world, led me to the interpretation that Trump is responsible, at least partly, for increase in racist tweets.
Trump's insistence on saying "Chinese virus" is normalizing and encouraging racism. Here is the data.
I plotted the number of tweets & replies (excluding RT) that use specific racial slurs in the past week, and saw an increase after Trump first tweeted "Chinese virus" (3/17 7p).
When the Commander-in-Chief is spewing out racist, xenophobic BS and refusing to back down, people are of course emboldened to do the same, both on Twitter and in the real world.
I used Twitter's Search API @TwitterDev to search for tweets that contain the slurs. I did not include retweeted or quoted tweets, as the RT may not be endorsing the original tweet. So I'm potentially under-counting.