The Datafied State
Starting points, questions, and invitations for a new research agenda
The Datafied State is one remade by the data sources and infrastructures, computational tools and techniques that are being adopted across Government just as they are in the private sector. There is not an absolute distinction between public and private sectors in the Datafied State but more of a blurred boundary. Government does not exist separately and outside of the tech industry, and its relationship to this industry is not simply as a source of regulatory pressure. Governments also procure, develop, implement, and legally mandate the use of digital and computational systems. Government use of tech — and the transformation of government through its use — is the primary interest and concern of Data & Society’s research on the Datafied State.
We are sharing a new Data & Society research agenda in the earliest stages of discussion and planning. Our purpose is to build a coalition of researchers and learn about similar projects, to share approaches, methods, and findings. At Data & Society, the Datafied State isn’t a single project, but a broad agenda that touches on all of our existing programs, builds on past research such as our study of the US Census Bureau and recent report on Electronic Visit Verification, and also suggests areas to expand and grow into. Here we describe the Datafied State and explain why we think studying it is important right now. This research agenda was drafted in collaboration with Postdoctoral Scholar Ranjit Singh who has studied the Aadhaar national ID system in India. We are working toward a broad global view of the State beyond Eurocentric preoccupations and assumptions. Below are some of our points of departure and the questions we’re obsessed with — questions we think need to be answered before we can move forward with recommendations or interventions. Do you share an interest in any of these questions? If so, please get in touch. — Jenna Burrell, Director of Research
The Datafied State arises from an abundance of data, in new forms and formats, some of it sought, accessed, and obtained through murky practices. In the United States, the drive toward a Datafied State is often shaped by institutional crisis. Understaffing and budget cuts combined with rising demand for services produce unmanageable caseloads. New tech offers hope to government offices and agencies for ways of organizing and automating data to streamline workflows. Elsewhere in the world, it is propelled forward by visions of becoming modern. Globally, the push for data and computational algorithms is also shaped by questions of legitimacy; concern that agents of the State are biased or may even commit fraud.
How datafied is the State today? How algorithmic? How automated? How can we find out?
The promise of tech in the public interest
Approaches to computational data-driven systems within government hold out the possibility for tech built in the “public interest.” Governments seek to solve problems and prioritize values beyond market fit and return on investment. Such work offers an in-house opportunity to model responsible, accountable, and accessible tech. In the United States, mail delivery guided by a universal access mandate has inspired pilot projects that begin with the challenge of rural mail delivery rather than treating rural mail service recipients as an afterthought. Governments serve citizens, constituents, the public, rather than customers, and they interface with advocacy groups, unions, and not just individuals. This move from a consumer focus to what is in the public interest is a consequential reframing.
An opportunity to peer inside the black box
Mandates for public accountability and the stewardship of tax revenue also mean that there is greater potential for researchers to peer into the workings of a Datafied State. There is a paper trail and legal requirements to provide access to public records. There is still a lot we don’t know, but there’s the possibility of knowing more with the right research framing and questions.
In what ways is the Datafied State new?
Reading work on computational tools in government over the decades, things start to look suspiciously like a cut-and-paste job. “Actuarial instruments” become “expert systems” which become “decision-support systems” and then “algorithms” which are sometimes specified as “machine learning” and, increasingly, as “artificial intelligence.” It is not always clear how much this evolving conversation is about changes in technology and how much is rebranding, leveraging the currency of new terminology.
For some historical perspective, consider Max Weber writing about bureaucracy as an innovation that emerged hand-in-hand with the Industrial Revolution. He describes the rise of the expert functionary who follows “calculable rules” and decides “without regard for persons,” displacing rule by feudal lords who were “moved by personal sympathy and favor, by grace and gratitude.” The case he made in 1922 is strikingly similar to the arguments now made for computational tools to substitute for the judgment of bureaucrats.
To collect data on populations and discern trends, to calculate and score risk, this is not a new trend in government. Neither is the use of checklists, weighted scoring systems, or other practices aimed at streamlining, standardizing, and depersonalizing the process.
So what is new here? Is it the new types of data (e.g., “data exhaust” or “behavioral trace data”) captured incidentally from life rather than composed for databases? Is it a matter of magnitude, pervasiveness, or complexity? Are new algorithmic tools closer to serving as a substitute for human intelligence as some proponents claim?
In what ways is it an Algorithmic State?
Essential to the Datafied State are the techniques used to process data. What are the characteristics of newer algorithms? What is the social life of these algorithms? There are new and distinct challenges presented by the opacity of contemporary computational tools, such as “AI.” A report by a team at Stanford Law School laments that federal agencies in the US are often using less “sophisticated” AI.
But what exactly does “more sophistication” yield?
We know that new techniques also have downsides. Their massive computational demands lead to both the centralization of power (to those with access to the requisite infrastructure) and a proportionately larger carbon footprint.
In what ways is it an Automated State?
The automated state is one that seeks to replace human workers with machines. There are three general motivations. One, the desire to leverage computational speed to handle rote and routinized work more efficiently. Two, the desire to improve the accuracy, fairness, or consistency of decision-making in light of human fallibility. Three, the desire to depoliticize decision-making (or appear to) by placing it out of reach of human discretion. These motivations, however, raise distinctive concerns about oversight and accountability and about the ability to seek recourse in the case of errors or bugs in decision-making. Efforts to automate have implications for participatory democracy. An algorithmic state can lead to automation bias where human decision-makers trust computational tools and their recommendations more than they should, but it can also be implemented in ways that exclude human intervention entirely. EU legislation, with the GDPR as a prime case, has emphasized the right to a human in the loop.
What does it make sense to automate in government? In what contexts is human-AI partnership most critical?
In what ways is it a Surveillance State?
The new and more widespread use of identification techniques — facial recognition and other types of biometrics in particular — raises fundamental concerns about established civil liberties and universal human rights. For those targeted or supervised by the State, the way the Datafied State operates as a surveillance state is very apparent, while it may be less obvious to others. This has implications for inequality since those who escape this scrutiny may benefit from it or may conveniently ignore it. By considering this frame, we focus attention on how datafication processes reconfigure power and control. State power is expanded through the widening net of surveillance and the use of tools of automated detection and enforcement. But this is experienced unevenly.
How does the State discriminate? How does such a technological regime function to oppress non-dominant groups in new or different ways? How do citizens and groups resist? How does the drive to feed data to algorithms or the ease of acquiring data end up altering the way the State functions, violating or operating in a gray area of civil liberties? What guidelines for data acquisition and use by the State could achieve the protection of civil liberties and due process?
How does the Datafied State Emerge Globally?
Emerging investments in digitalization and datafication in many parts of the world follow a developmentalist logic and consequently are pursued as an attempt to “catch up” to advanced capitalist societies by appropriating data-driven interventions as exemplars for what it means to be modern and efficient. Yet, the origins and evolution of data practices of the State are multiple and do not always follow assumed or historic patterns of technology transfer from north to south or from the west to rest. Cash transfer programs globally have been modeled on the success of the Bolsa Familia program in Brazil. The proliferation of biometrics-based identification systems has similarly a lot to do with the success of Aadhaar in India. Financial inclusion through mobile phones borrows inspiration from M-Pesa in Kenya.
How does the Datafied State represent the possibility of being modern? Where are the centers of influence, reinvention, or advancement? How does geopolitics shape national policies and practices surrounding datafication?
The policy discourse around India’s development, for example, has shifted over time from investing in bare necessities for subsistence (food, clothing, and housing) to material infrastructures (for electricity, roads, water) to, ultimately, the infrastructure for data-driven services (drawing together bank accounts, biometric-based identities, and mobile phone numbers). The utopian vision of the Indian state sets its ambition well beyond “catching up” with the West. Government leaders and rank-and-file workers building these systems aim to reinvent the oversight and management of citizens and the State through pervasive digitalization. Plans, of course, have a way of falling short of reality, but in the pursuit of a certain vision, even unsuccessful plans may reshape government and society in lasting ways.
In what ways is the Datafied State experienced?
The reorganization of the State via datafication changes the way in which bureaucratic procedures are navigated and experienced on an everyday basis. Anthropologists of the State have long been interested in the uneven consequences of bureaucratic procedures that mediate state-citizen relations. For example, consider the work undertaken by low-resolution citizens in India seeking (often unsuccessfully) to navigate the national Aadhaar biometric identification system which now mediates many benefits programs. There are life-or-death consequences of failing to be recognized within this system. If citizenship is a practical accomplishment of how the State is organized, then the Datafied State fundamentally changes the nature of exercising and experiencing citizenship. In a Datafied State, a citizen without data becomes less of a citizen. This raises a myriad of questions.
How do citizens secure and claim representation in accordance with the core data categories used to organize government services? Or conversely, how do they strategically disappear from data systems that constitute their relationship to the State?
In what ways is the Datafied State aspirational?
There are questions about what is genuinely new here, but also about the present reality of the Datafied State.
What projects are merely proposals, experiments, or exploratory ideas and which are operating components of how the State functions? How are algorithmic tools actually incorporated into work practices within government? Which proposals are projections of futurism, either utopian or dystopian? What is rooted in hope? What is a figment of hype?
For example, the United States Postal Service has pursued two distinct autonomous vehicle projects, one for rural service and one for long-haul mail transport. Commercial efforts to develop fully self-driving vehicles have been underway for over a decade, but some programs have been abandoned while others have been chastised for overpromising on capabilities and timeline.
Where do we place experimental imaginings (perhaps never to be realized) in our understanding of the Datafied State? What purpose does the aspirational serve even if it is never realized? How do these visions relate to the sometimes mundane present state of algorithms in government?
Ambitions
In the coming months we will be working to identify important subtopics and focus on particular cases of data-centric technologies in government that could help us answer our questions about the Datafied State. In addition to finding new projects, we plan to support research and writing by others in our extended community who are also tackling this topic. If this research community is successful:
- We will better understand public administration and how new digital tech and datafication processes are altering this work.
- We will have one or more models of how tech development and use could be guided by public interest values such as transparency requirements, due process, and equitability rather than private sector interests like profit maximization and the drive to scale at any cost.
- We will gain insight into how to guide procurement and development decisions within government toward the public interest.
- We will better understand new threats to civil liberties that can emerge in novel applications of digital tech by the ever more powerful Datafied State.
- We will advance the conversation about how existing legal structures hold up against the threats to civil liberties or whether new ones are needed and what form they should take.
- We will show how inequality can be propagated or (alternatively) checked by the digital and data infrastructures built or used by the State.
- We will produce well-grounded training and educational materials to serve government workers interested in deploying digital tech aligned with public interest values and with an awareness of the social and ethical implications of their use.
If your work connects to Data & Society’s developing research agenda on the Datafied State, please get in touch.