Ethnic Minority Legibility in China

• In Event: Bureaucracy, Institutions, and Authoritarian Rule

Abstract

Although scholars have established that minorities present a problem of legibility for authoritarian regimes (Dimitrov and Sassoon 2014; Blaydes 2018; Blaydes 2020), the specific techniques through which this obstacle can be overcome necessitate further theoretical development and empirical testing.
This paper argues that authoritarian regimes may pursue two different strategies for penetrating minorities and collecting the information they need to improve legibility. The first is building up human informant networks, whereas the second is the use of artificial intelligence (AI) to engage in technologically sophisticated surveillance. There are tradeoffs to each strategy. Although very costly to build and operate, human intelligence provides superior information about latent minority discontent that can be used to deploy repression selectively and effectively. In turn, the AI surveillance infrastructure can be built quickly and at a relatively low cost, but its capacity to provide actionable intelligence in the absence of a strong human informant network is limited. Instead of selective repression, AI facilitates mass incarceration due to its reliance on crude algorithms that link specific behaviors by minority members with subversive intent.
These arguments are tested through empirical data from China that allow us to shed light on the management of territorially concentrated minority groups like the Tibetans and the Uighurs. The paper demonstrates that the incapacity of the regime to build strong human collection networks in Xinjiang and Tibet justified the adoption of aggressive AI surveillance. The paper concludes that although AI generates massive amounts of data, it facilitates totalistic coercion rather than the sophisticate targeted repression that can be deployed when levels of legibility are high. The paper is based on a large corpus of internal circulation (neibu) materials from China.

Author

• Martin Dimitrov, Tulane University