Biometric Authentication and the Ontological Erosion of Self-Identity — Epoche C1
The concrete change, and three things the word 'identity' is doing When a Korean bank replaces the thirteen-digit resident registration number as its means of confirming who a customer is with a fingerprint read from that customer's hand, the change looks administrative: one credential swapped for a more convenient one. The claim of this essay is that it is not administrative, and that the reason is missed by the debate as usually conducted. Public discussion in South Korea, shaped by a long series of large-scale leaks, has treated biometrics as a data-protection question — how securely is the fingerprint stored, who may see it, what happens if it escapes. That question is real and is not the deepest one. The deeper one is what happens to the concept of a person's identity when the thing that settles it is a measurement of their body, taken and adjudicated by someone else. Answering that requires separating three senses of 'identity' that the compressed version of this argument ran together, and whose conflation is the reason its conclusion sounded either alarming or empty depending on the reader. Identification : determining which of many enrolled persons someone is — a one-to-many search. Authentication , or verification: confirming a claim to be one particular enrolled person — a one-to-one comparison. This is what a phone's fingerprint sensor does. Personal identity in the philosophical sense: what makes a person at one time the same person as at an earlier time. This is a question about persistence, not about records. John Locke gave the modern debate its shape in the chapter 'Of Identity and Diversity' added to the second edition of his Essay Concerning Human Understanding (1694), where he argued that sameness of person consists in continuity of consciousness rather than in sameness of body or soul — a person is the same person as the one whose past actions they can appropriate as their own. To these must be added a fourth, less technical sense: identity as a person's account of who they are — their history, commitments and relationships. The received view, which this essay is testing, is that biometrics touches only the first two, and that the philosophical and personal senses are safely insulated. The argument below is that the insulation fails, and that the failure has a specific mechanism. What a biometric system actually computes The mechanism cannot be seen without knowing what the machine does, and here the compressed version of this essay asserted something false that ought to be corrected before anything is built on it. It said that the individual's physical self 'becomes the data'. It does not. What is enrolled is not a fingerprint but a template : a fixed-length vector of features extracted from an image — for fingerprints, typically the positions and orientations of ridge endings and bifurcations. The original image is normally discarded. A template is a lossy derivative, and two templates from the same finger, captured minutes apart, are never identical, because pressure, moisture and placement differ. It follows that a biometric system cannot compare for equality. It computes a similarity score $s(x, t)$ between a fresh sample $x$ and a stored template $t$, and declares a match when $$s(x, t) \ge \tau$$ for a threshold $\tau$ chosen by whoever configures the system. Two error rates follow immediately and cannot be independently minimised. The false match rate $\mathrm{FMR}(\tau)$ is the probability that samples from different people score at or above $\tau$; the false non-match rate $\mathrm{FNMR}(\tau)$ is the probability that two samples from the same person do not. Lowering $\tau$ admits more impostors and rejects fewer genuine users; raising it does the reverse. There is no setting at which both vanish, because the score distributions of genuine and impostor pairs overlap. This is the technical fact on which the whole ethical argument turns, and it is stronger than the claim it replaces. The system does not store your identity and does not assert it. It asserts that a number crossed a line drawn by an administrator. Where a password answers a question with a determinate yes or no — you either know the string or you do not — a biometric answers a question of degree, and someone else chooses how much degree is enough. Agre's capture model, correctly stated The compressed essay invoked Philip Agre's 1994 paper 'Surveillance and Capture: Two Models of Privacy' for the proposition that turning a biological feature into machine-readable data alters a person's relation to their identity. That is not what Agre argued, and the misattribution is worth repairing because his actual thesis is more useful here. Agre distinguished two models of how information systems bear on persons. The surveillance model is visual and political: a watcher, a watched, a record kept without consent. The capture model is linguistic and organisational: a system imposes a 'grammar of action', a formal vocabulary of permitted steps, and human activity is then reorganised so that it can be expressed in that grammar. Capture, on Agre's account, does not merely record activity; it requires the activity to be restructured in advance so as to be recordable. Biometric authentication is a clear case of capture in exactly that sense, and saying so grounds a claim the essay previously left floating. The body is disciplined into machine-legible postures: present the finger flat and still, hold the face square to the camera under adequate light, remove the mask, do not blink, look at the dot. A grammar of bodily action is imposed, and access is conditional on performing it correctly. That is a real alteration in the relation between a person and the system, and it is prior to any question of leakage. Where the philosophical sense of identity is actually touched The step from capture to the erosion of self-identity needs an intermediate term, and the most precise one comes from Paul Ricoeur's Oneself as Anothe