Since Xi Jinping consolidated his power in China in 2013, he has taken the concept of ‘harmonious society’ (和谐社会) to the next level: zero tolerance for public expression of grievances of any kind and by any social or ethnic group, and widespread use of state violence to crush anyone daring to raise their voices. This includes anything from AI-assisted online censorship, imposed face-recognition, house arrest, arrests and threats to families, disappearance, beatings in public spaces, torture, rape, incarceration in camps, starvation, and killing. The latter is particularly true for ethnic minorities, and mostly for Uyghurs.
For more, read Global Voices’ Special Coverage on Documenting state-directed persecution of minorities in China's Xinjiang region
Yet despite the immense danger of voicing dissent, Chinese citizens of various social background, age groups, ethnicities continue to resist. One major recent example is the White Paper movement that in November 2022 saw people protest against inhumane ‘zero-Covid’ policies leading among other things to starvation or children burning to death in locked buildings. Xi eventually reversed the extreme policy after citizens demonstrated across China including in rich and conservative urban centers such as Shanghai.
To have a sense of the scale and nature of dissent in China today, Global Voices talked to Taiwan-based Kevin Slaten, who is a Program Manager for Asia at Freedom House, a US-based think tank. Slaten, who tweets here, heads the China Dissent Monitor (CDM), which has mapped more than 1,500 instances of dissent across China that combines data collected by AI and human checkers to produce a representative sample of civic protest in China today. It is accessible here.
The interview took place in English in a café in Taipei and was edited for style and brevity.
Filip Noubel (FN): Can you describe the methodology for the CDM and particularly how you combine human and AI research and analysis methods to produce such a comprehensive body of work, given the size of China?
Kevin Slaten (KS): The impetus for the CDM is an information gap which results from two factors: general censorship in China, which is constantly updated to prevent certain information from being shared widely. Indeed information about protest is not easy to find, thus there is no central repository. The second factor is that people who went through the trouble of finding this information in China have been arrested for doing so. The criminalization of recording and sharing images of protest has thus increased the information gap because Chinese people were in the best position to monitor dissent.
CDM addresses this gap, and we do it outside of China for safety reasons. We collect data into three streams: independent research (news reports mostly, and social media inside and outside of China), weibo scrapping and analysis through machine learning (in collaboration with the Taiwan-based DoubleThink Lab to catch the posts missed in the first stream due to the censorship that is built in Chinese search engines), and NGO partners who work with specific communities and can access what is not present online. In the end, we put the three streams together in a combination of AI and human work, and code them for comparability. We use AI to enhance what people are doing, which is the best way to use it: The computer gets rid of what we don't need as we train it. This is a good example showing AI can be used for pro-human rights purposes.
We also have three types of verification: the first tier has visually verifiable evidence (photo and videos), the second has more than one non-visual source per item, and the last tier has only one source.
Our public and coded database offers about 1,500 events for the past 12 months. We acknowledge that we are missing things, as for example, protests in the countryside: Years ago, the Chinese government would publish its own data and the figure would be of 100,000 “mass incidents” per year. So CDM is a piece of an entire universe, but currently it is the most comprehensive database available.
FN: Housing protests account for about one-third of your data. Could you explain why?
KS: This includes three different groups: home owners, home buyers who face homes not being constructed or handed over, and shop buyers. They are all linked because they face the same problem: Real-estate companies do not have the cash flow to finish the construction because there are regulations that allow to take money from one project and inject it into another one. This national-scale pyramid scheme makes for a large part of China's GDP. There was a tightening of regulations a few years ago, but as constructions slowed down in part because of COVID, regulations have been loosened again and companies recapitalized by the state.
This being said, our data doesn't show any major decrease in housing-related protests. How long can this discontent go on before people start organizing themselves – something we witnesses last summer? We saw mortgage protests when people refused to pay mortgage because their homes were not being built, or not built entirely. Our research shows that the government cracks down violently on such protests, perhaps because they tend to be larger in numbers of participants compared to other protest so they might appear as more threatening. Besides, such protesters often blame the government for lack of oversight on companies, and also expect the government to fix the situation.
FN: Can you explain why risks for and violence against non-Han protestors differ from what the Han majority experiences?
KS: Information about those groups is even more of a snapshot because we mostly get data from NGOs, and their access is also limited. The risk is that we undercount protests as non-Hans are more likely not to post anything because of the risks to their security. There are now very sophisticated systems of off- and online control in China, and those used against non-Han groups are more severe: they include completely shutting the internet in entire cities, in-person surveillance, scanning of phones, and for the Uyghurs, imprisoning people to terrorize them. But I would add religious minorities here, and that also includes Han people.
FN: Overall what would you say are the main changes in strategy, tools, and impact shaping dissent in China today – both from the protestors’ side and the government's side?
KS: One of the new things we see is more automation on the side of repression: Algorithms are twisted to have “positive news” only, and to downplay news that the government doesn't want people to notice. On the dissent side, we notice the occurrence of non-centralized movements. Here symbolism plays a key role: people connect mostly simply by observing and modelling, without necessarily having direct communication. Borrowing slogans and expressing solidarity is often enough to inspire dissent. What our data shows is that there is definitely protest and dissent happening in China, despite the very high risks for the people involved.
The White Paper protest, which we see as an anti-lockdown protest, includes at least 200 related events opposing the ‘zero-Covid’ policy over several months. It is clear that people participating in those protests were aware of other groups and protests. The main question today is whether the memory of the protest will it have a shadow in the coming years. Chinese people saw that their government did listen and eventually amended its ‘zero-Covid’ policy. That's very empowering and came as the result of a totally decentralized movement.