Veil of ignorance

Working Definition

The veil of ignorance is a procedural test of fairness for social and institutional rules: a principle is just if and only if it would be agreed upon by rational parties who do not know what position they will occupy once the veil is lifted—neither their class, natural talents, sex, race, or generation, nor their personal conception of the good life. In John Rawls’s philosophy, the veil defines the “original position,” a hypothetical social-contract baseline designed to extract impartial principles from self-interested reasoning without assuming benevolence: because I do not know which citizen I will turn out to be, I must protect the fundamental interests of every position. In everyday usage the phrase means “design the rule as if you did not know which side of it you will land on”; in technical political philosophy and decision theory, everything hinges on what information the veil blocks and what decision rule the parties use under uncertainty.

Genesis and Global Lineage

John Rawls developed the construction from the late 1950s (Justice as Fairness, 1958) and gave it canonical form in A Theory of Justice (1971), deriving two lexically ordered principles from choice behind the veil: equal basic liberties for all citizens, and the “difference principle,” under which social and economic inequalities are permissible only if they work to the maximum benefit of the least-advantaged members of society. The device revitalized the social-contract tradition of Locke, Rousseau, and Kant against the utilitarianism that had dominated Anglo-American philosophy since Bentham and Sidgwick. Crucially, the formal model of an ignorant chooser had already been invented inside welfare economics: William Vickrey (1945) and John C. Harsanyi (1953, 1955) modeled an impartial observer who does not know which individual they will be and deduced average utilitarianism. Rawls adopted the setup but thickened the veil—stripping away probability distributions and personal utility functions—and replaced Bayesian expected-utility maximization with the maximin rule (maximize the minimum outcome). In Political Liberalism (1993) and Justice as Fairness: A Restatement (2001), Rawls recast the original position as a heuristic “device of representation” for constitutional democracies.

Central European Lineage

The welfare-economics precursor and sharpest decision-theoretic critic of Rawls’s veil of ignorance was the Hungarian-born Nobel laureate John C. Harsanyi (János Harsányi), intellectually formed in Budapest. Educated at the Lutheran Fasori Gimnázium (which also produced John von Neumann and Eugene Wigner), Harsanyi earned a pharmacy degree (1944) and a doctorate in philosophy and sociology (1947) at the University of Budapest, teaching at its Institute of Sociology in 1947–1948 until forced out for his anti-Marxist views. After escaping Hungary in April 1950 and emigrating to Australia and the US, Harsanyi published two landmark papers in the Journal of Political Economy (1953, 1955). He argued that social-welfare judgments become impartial when made in “complete ignorance of what the chooser’s own relative position (and that of his near-by social group) would be within the system chosen,” assigning an equal probability (1/n) to occupying any person’s place. Rawls cited Harsanyi’s 1953 and 1955 papers in A Theory of Justice while rejecting their utilitarian conclusion. In 1975 Harsanyi answered with a classic critique in the American Political Science Review (Can the Maximin Principle Serve as a Basis for Morality?), showing that under Bayesian rationality axioms, ignorance of one’s identity mandates maximizing average expected utility rather than Rawls’s extreme risk aversion (maximin).

Why Now

The fiftieth anniversary of A Theory of Justice in 2021 coincided with institutional problems that brought the veil of ignorance back as an operational design protocol. First, during the COVID-19 pandemic (2020–2021), bioethics committees designing ventilator triage and vaccine-priority schedules pitted utilitarian life-years maximization against Rawlsian maximin priority for the most vulnerable. Second, in AI alignment: in a 2023 PNAS study (Using the Veil of Ignorance to align AI systems with principles of justice), Google DeepMind researchers placed human participants behind a veil of ignorance when choosing distributive rules for an AI assistant; participants deliberating without knowing their own handicap consistently endorsed rules protecting the worst-off player and kept that preference after their actual position was revealed. Third, in climate negotiations (the COP27 Loss and Damage Fund, 2022) and post-2020 meritocracy debates (revived by Michael Sandel), Rawls’s core claim—that birthplace and natural talents are morally arbitrary “accidents of birth”—returned as the primary philosophical counterweight to market desert.

Explanatory Power

The veil of ignorance separates two questions that political argument constantly conflates: “what rule favors people in my current position?” and “what rule is fair regardless of position?” First, in tax and social-insurance design: a healthcare or unemployment rule that a healthy, high-earning professional would endorse only while knowing their own medical and work history fails the veil test; behind the veil, social insurance is rational protection against the tail risk of disability. Second, in platform governance: instead of optimizing rules for the average user or loudest advertiser, the veil forces designers to inspect how an automated ban looks to the least powerful account holder falsely flagged by a classifier. Third, as a diagnostic of moral boundaries: asking who stands behind the veil—only citizens of one state (Rawls’s The Law of Peoples, 1999), or all humans globally (Thomas Pogge, Charles Beitz), future generations, and severely disabled persons (Martha Nussbaum)—shows that the hardest step is deciding the membership list of the original position before the veil drops.

Axes of Debate

Underappreciated Angle

The most counterintuitive implication of the veil of ignorance for the twenty-first century is that it makes deliberate ignorance a precondition of institutional justice. We live in an era of actuarial profiling, genomic screening, credit scoring, and predictive AI models whose engineering goal is to lift the veil of ignorance completely: an insurer that knows your polygenic risk score, an employer that predicts your burnout risk, and a platform that knows your reservation price no longer face uncertainty about who is who. Yet solidarity in health insurance, public education, and the rule of law historically depended on a natural veil of ignorance—we pooled risks precisely because none of us knew in advance who would get sick or whose skills would be automated away. Once predictive algorithms strip away that uncertainty, preserving fairness requires building synthetic veils of ignorance by law and design: blinding rules, banning certain predictive features, and withholding individuated data from decision-makers. Privacy and algorithmic-blinding laws are thus not merely protections for secrecy; they are the institutional infrastructure that keeps the social contract from unraveling when machines know too much.