I’m a researcher at the UC Berkeley Center for Long-Term Cybersecurity, where I direct the Daylight Lab. This newsletter is my work as I do it: more than half-baked, less than peer-reviewed.
The ability to provision systems of identity is the most important vector of power in the world. Identities build notions of personhood—persons over whom control can be exerted. Identities are logically prior to any notion of “legitimate” control. As such, the particular mechanics of these identities radically shape life chances.
Historically, state-issued identities have been our primary identifiers. These identities are mostly static, and primarily issued at birth. Their primary purpose is to constrain mobilities across national borders, mediate access to rights and opportunities (civic and private).
However, the private sector has created its own forms of identity. First, credit ratings; now, with the advent of networked technologies, surveillance capitalism. These identities are ever-morphing; they are identities made of data over time, condensed by algorithms into “profiles.” They are what Aneesh Aneesh calls system identities.1
System identities are the true revolution of the so-called Information age.2 They are the object of accumulation in surveillance capitalism; they produce identities amenable to capitalization through trade, sale, and use in targeted advertising.3 For states and tech companies alike, they are the heart of contemporary securitization.4
What makes identities meaningful?
Among competing identity systems—state-issued and otherwise—what makes particular systems meaningful? Why do we use some and not others? In other words, from what properties does a particular identity system derive its power?
Three properties make a particular system of identity amenable to participating in an “impersonal, anonymous society:”5
Like money, the value of identities derives in part from their scarcity. Identity systems have a particular constraint: within any one system, it is difficult to have more than one identity. For state identities, agencies like the DMV assure a one-to-one relationship between individuals and identifiers.6 In ML-driven system identities, algorithms collapse data-streams into trackable identities, which construct a profile of an individual over time.7
Identities’ value also derives from their difficulty to change. Names provide stable identifiers and clear rules by which those identifies can change. And system identities, while ever-morphing, can neither be altered in arbitrary ways nor entirely removed.8
Finally, identity systems’ value stems from their “fit” to particular notions of personhood. Existing identity systems describe atomic individuals, so-called “natural-born persons.” This design coheres to a particular notion of legal personhood, one with historic roots in Anglo-American capitalism, one amenable to capital accumulation. The key aspect of any identity system is that it coheres specifically to a popular notion of personhood in a way that provides instrumental value to whichever systems of exchange predominate.9
States of knowing
How does one make a system of identity that has these critical properties? Through epistemic power: the power to learn, “imposed by private commercial mechanisms of information capture, production, analysis, and sales.”10 The construction and maintenance of identities—state or system—is a fundamentally epistemic activity. Meaningful identities are born of epistemic power.
As we rethink violence, we must too rethink the state. Insofar as identities shape life chances, they enforce legitimate(d) violence. Insofar as they enforce legitimate(d) violence, they define states as such. To provision widely-meaningful identities is to legitimate violence, to exert state-like power.
I call these entities, whose ability to enforce legitimate(d) violence derives primarily from their epistemic power, states of knowing. Tech companies motivated by profit and states motivated by securitization, incentives align to cooperate to build comprehensive system identities. (The U.S. government’s chief complaint against Facebook is, in a word, that they fail to adequately cooperate in its regime of securitization).
Sam Altman wants to scan every eyeball in the world. His goal is, in effect, to provision an identity system. The whole project exudes big “white man’s burden” energy. The problem with identities is not that they exist, nor who provisions them per se. The problem is that there is no popular governance over their mechanisms or meaning.11
Sam Altman wants to scan every eyeball in the world. I want to empower people to scan their own eyeballs.
Some questions for you
What would popular governance over an identity system look like?
How could a grassroots movement produce a locally meaningful identity system at a municipal or popular assembly scale?
What kinds of services could a locally meaningful identity system provision? Are there any services it could not provision?
Philip Mirowski and Edward Nik-Khah. The knowledge we have lost in information: the history of information in modern economics. 2017.
Nick Couldry and Ulises A Mejias. The Costs of Connection. 2019.
The NSA inferred citizenship from data traces. “a legal citizen whose algorithmic foreignness measure crossed the 51% threshold of confidence could be treated as a foreign and thus denied the Fourth Amendment protection of privacy” (Burrell & Fourcade, 2020).
Contemporary securitization relies on system identities, often collected via relationships with tech companies. These relationships can be voluntary, like Palantir's, or involuntary, like AT&T's participation in Snowden-era mass surveillance.
Even in the private sector, data traces do the work of security. See Anne Jonas and Jenna Burrell. “Friction, snake oil, and weird countries: Cybersecurity systems could deepen global inequality through regional blocking.” Big Data & Society 6.1 (2019).
In the private sector, the use of system identities was presaged by credit scores. See Josh Lauer. Creditworthy. Columbia University Press. 2017.
Ernest Gellner, Nations and Nationalism. 1983.
Broadly, identifiers are authentication methods. The DMV issues an identifier that combines a possession factor—your license, which scans—and an inherence factor—a photo of your face, which should look like the holder.
Their identifier is a continuous, behavior-based sort: your ability to emit signals that look like you---and your inability to emit signals that look any different---authenticate you.
To alter state identities through immigration or bureaucratic processes requires tremendous time, effort, skill, social power, and luck. To alter system identities purposefully requires tremendous technical skill (e.g., of cryptography and statistical approaches to privacy). After all, the process that generates them—you, and your behavior—is difficult to change.
Liability is a particular kind of securitization. In liberal societies, liability (assigning blame to a legal notion of personhood, natural or corporate) is a prerequisite for smoothly-running systems of exchange. If those systems of exchange break down, so does the state.
In other words, liberal societies are as functional as their systems of liability, which are in turn built “on top of” specific systems of identity.
Shoshana Zuboff. “The coup we are not talking about.” The New York Times 29 (2021).
Privacy regulations affecting system identities are partial and incomplete. And there has never been meaningful popular oversight over state identity systems.