Politics of Sociotechnical Systems
Politics of data in the public sector
The State’s Politics of Fake Data. in progress. (with danah boyd)
States wield symbolic power to legitimize the data they produce, utilize, and disseminate. In turn, state data acts to affirm the legitimacy of the state itself. Within this dynamic, much scholarly attention has been directed at the importance of genuine data production, emphasizing its representational accuracy. However, the meaning and production of 'fake data,' commonly perceived as either technically erroneous (wrong data) or normatively undesirable (bad data), have been somewhat overlooked. This paper challenges this predominantly representationalist perspective on fake data, drawing on ethnographic, interview, and archival data from two distinct cases where fake data controversies arose. The first delves into the methods employed by Chinese bureaucrats in documenting neighborhood issues and activities. The second scrutinizes a specific facet of the United States Census Bureau's census data production. Rather than accepting the notion of fake data as an inherent, independent, and immutable entity, we encourage scholars to transcend debates centered on data's representational aspects. Instead, we advocate for an exploration of how people do things with fake data throughout its social trajectories. In mapping the ways states conceptualize, interpret, and navigate fake data, we highlight how—from its inception to its rectification and interpretation—multiple actors shape and redefine it at every phase, imbuing it with diverse meanings that often transcend mere representation of reality. Consequently, designating data as 'fake' goes beyond mere acknowledgment of its representational shortcomings. It's a sociopolitical act that demands a nuanced, context-rich analysis.
Making Data-Driven by Driving Human. in progress. (with Canhui Liu)
Chinese Social Credit Systems
Governing Everything: The Sociotechnical Life of a Chinese Social Credit System (book manuscript in progress)
A Tale of Two Social Credit Systems: The Succeeded and Failed Adoption of Machine Learning in Sociotechnical Infrastructures. Oxford Handbook of the Sociology of Machine Learning, edited by Christian Borch and Juan Pablo Pardo-Guerra. Oxford: University of Oxford Press |download|
Machine learning technologies have permeated diverse sectors, catalyzing transformative shifts in the understanding, management, and navigation of complex sociotechnical systems. However, how are machine learning technologies adopted in different scenarios, and what are the necessary sociotechnical conditions? This chapter undertakes a comparative analysis of machine learning technologies adoption in two Chinese social credit systems. The central argument of this chapter revolves around two primary components: diverse data input and well-defined outcomes. Both elements are fundamental to the effective deployment of machine learning models and influence their accuracy, relevance, and utility. The success or failure of machine learning adoption is not solely a technical or social matter. Instead, as the chapter underscores, there is an intricate balance between technical prowess and social compatibility, both of which are indispensable for successful technology adoption.
Pervasive surveillance in modern society has raised mounting debates, which are largely concentrated on the ethical dimension and lack sociological examination. Drawing on innovative national survey data, this study analyzes public opinion about social credit systems (SCSs), an emerging infrastructure that expands the depth and breadth of surveillance in China. I find a general high support for expanding surveillance and punishment yet key variations among different social groups. Counterintuitively, people with higher political capital do not wholly embrace the expanding surveillance and punishment. For example, Chinese Communist Party members are less likely to support state-centered SCSs compared to the general public. Higher political trust in the regime and socioeconomic status is consistently correlated with higher support, while different media consumption showed limited correlations. This study proposes an alternative theorization of surveillance and enriches our understanding of the heterogeneity and dynamic of the state and public in the authoritarian regime.
In 2014, the Chinese government proposed to build a social credit system (SCS) to better collect and evaluate citizens’ creditworthiness, and grant rewards and punishments based on one’s social credit. Since then, various SCS pilots have been enacted. While current media and scholars often perceive SCS as a single and unified system, this paper argues that there are in fact multiple SCSs in China. I identify four main types of SCS and articulate the relationships among them. Each SCS has different assumptions, operationalizations, and implementations. China's central bank, People's Bank of China and the macroeconomic management agency National Development and Reform Commission are the two most important actors in the design and implementation of the multiple SCSs. Yet their distinctive views about what a "credit" is and what an SCS should be produced great tensions on the SCS landscape. I also historize current SCSs and show that many elements and assumptions of SCSs can be traced back to a broader People’s Republic of China’s (PRC) political history. At last, I propose an alternative theoretical framework to understand Chinese SCSs as a symbolic system with performative power that is more than a simple repressive and direct political project.