From Census to Sensors: The Old Data Regime and the Data Revolution

• In Event: Market Codes: Bureaucrats, Brokers, and Algorithms

Abstract

This article examines a fundamental transformation in the relationship between social scientific knowledge and state governance through the lens of two “data regimes.” The earlier bureaucratic regime was characterized by governmental agencies that supplied most data and conducted analysis within a framework of bureaucratic epistemology — expensive, theory-driven, and professionally gatekept. This system, built on representative surveys and official reporting mechanisms, made society legible to the state and supported its commitment to rational governance. The techniques that constitute this data regime persist, and still dominate in some domains, but they are no longer hegemonic. A new set of techniques for knowing about populations has grown up alongside them.

This new regime, which we call the brokerage regime, is characterized by a computational epistemology in which data production is cheap, decentralized, inductive, and dominated by private actors. This transformation has been driven by the erosion of trust in official statistics, increased politicization of government statistics, and the rise of a big data industry largely built on publicly funded information. In this new landscape, passive data collection through sensors and digital traces is slowly displacing expensive representative sampling, while algorithmic pattern detection is increasingly supplanting theory-guided analysis. Sensitive government data once requiring special protections is being recast as raw material for AI engines, in an ostensible effort to automate and optimize government operations.

What becomes of the state in the process? On the one hand, it may lose its capacity to collect information and allocate resources toward public goals, ceding sovereignty over critical functions to a patchwork of private entities. On the other hand, this hollowing out may mask an expansion of the state’s surveillance and police powers, supercharged by the production of increasingly precise “truths” about citizens and corporate entities that can be repurposed for law enforcement, social control, and economic extraction. We see and show evidence for both processes in recent developments in the United States.

Authors

• Adam Visokay, University of Washington

• Marion Fourcade, University of California-Berkeley

• Tyler McCormick, University of Washington

• Sarah Quinn, University of Washington