Reinterpreting the Rationality of Rawls's Veil of Ignorance — Epoche C1
The veil of ignorance is a stipulation about what the parties to Rawls's hypothetical agreement are not allowed to know. In A Theory of Justice (1971) the parties choose principles to govern the basic structure of their society — its constitution, its property regime, its economy — while ignorant of their own class, race, sex, natural talents, psychological propensities including their attitude to risk, and the content of their own conception of the good. The standing objection is that this device does not generate egalitarian principles so much as presuppose them: strip away everything that could distinguish one chooser from another and equal shares are the only thing left to choose, so the veil smuggles in a prior commitment to equality and dresses it as a derivation. The compressed version of this essay defended Rawls against that charge by describing the original position as a "pre-preferential" state, a moment prior to the formation of desires, and by reading the parties' choice as chronologically antecedent to the individuation of persons. That defence has to be withdrawn. It contradicts Rawls's own statements about what the parties possess, and the description it offers is, almost word for word, the characterisation that Michael Sandel used in 1982 to argue that the theory is incoherent. A better defence is available, and reaching it requires seeing precisely where the egalitarian conclusion enters — which turns out to be somewhere the objection does not look. What the veil removes, and what it leaves in place The parties behind the veil are not blanks. Rawls equips them with a great deal, and the list matters for everything that follows. They know all general facts: economics, psychology, sociology, the laws of nature. What they lack is particular information about themselves. They are rational in a deliberately narrow, instrumental sense — they take effective means to their ends and rank options coherently — and mutually disinterested , meaning each seeks to advance his own interests and takes no interest, benevolent or malevolent, in the interests of others. They know that they each have some rational plan of life, some conception of what is worth pursuing, even though they do not know its content. They therefore have determinate interests: each wants more rather than fewer of the primary goods — rights and liberties, powers and opportunities, income and wealth, and the social bases of self-respect — because those are useful whatever one's plan turns out to be. They know the circumstances of justice obtain: moderate scarcity, and a plurality of persons with differing aims. So the parties have preferences and pursue them. What they do not have is knowledge of whose preferences they are. This is a considerably weaker claim than the one the original essay made on Rawls's behalf, and Rawls needs the weaker one: parties with no interests at all could not choose anything, and a choice is what the argument requires. One item on the list of excluded knowledge deserves attention because the objection often turns on it. Rawls withholds knowledge of the parties' attitude to risk, and he is explicit that his argument does not proceed by attributing to them any particular such attitude. The argument is not "the parties are cautious, therefore they choose cautiously". If it were, the objection would be immediate: caution is a psychological trait like any other, and building it in would be exactly the smuggling alleged. The choice problem, made explicit Rawls's claim is that the parties would adopt the maximin rule: among the available options, choose the one whose worst possible outcome is best. Set against the natural alternative, this is where the whole dispute lives, and it can be displayed with three positions and two candidate social arrangements. Let the numbers be indices of how well a representative person in each position fares. Arrangement Position 1 Position 2 Position 3 Average Minimum A 1 8 9 6 1 B 4 5 6 5 4 Maximin compares the last column and selects B, since $4 > 1$. Now suppose instead that a party, not knowing which position he will occupy, treats the three as equally likely and maximises expected value in the manner standard in decision theory since von Neumann and Morgenstern. His calculation is $$\mathbb{E}[u] = \tfrac13(1) + \tfrac13(8) + \tfrac13(9) = 6 \quad \text{for A}, \qquad \tfrac13(4)+\tfrac13(5)+\tfrac13(6) = 5 \quad \text{for B},$$ and he selects A. Generalising, with $n$ positions each of probability $1/n$, expected utility is $\frac{1}{n}\sum_{i} u_i$, and maximising it is the same as maximising $\sum_i u_i$, which is to say maximising average utility. This is not a hostile reconstruction. John Harsanyi had derived it before Rawls wrote, in his 1955 paper on cardinal welfare, using an "impartial observer" who imagines an equal chance of occupying any person's position; and in 1975 he turned it directly against Rawls, arguing that maximin is not a rational decision rule at all. His illustration is memorable: a person who must choose between a better job in another city and a worse one at home, and who reasons by maximin, must refuse the better job because the aeroplane might crash — and, pressed consistently, must never cross a street. Maximin ignores the magnitude of the probabilities entirely, attending only to the worst case however remote. The same veil, then, yields utilitarianism in Harsanyi's hands and the difference principle in Rawls's. The veil is not what settles the matter, and any defence of Rawls that rests on the veil alone is defending the wrong thing. Where the egalitarian commitment actually enters Rawls knew this and stated the conditions under which maximin is a reasonable rule. There are three, given in A Theory of Justice §26, and each is a substantive claim rather than a formal one. No reliable basis for probabilities. The parties cannot assign likelihoods to occupying one position rather than another. Rawls specifically refuses the pri