Ex Ante Contractualism and Fairness.

Contractualism holds that an action is permissible only if some principle licensing it is justifiable to each individual. When outcomes are uncertain, ex ante contractualism looks to each individual’s prospects before it is known which outcome will be actual. By contrast, ex post contractualism assesses the distribution of outcomes.

Johann Frick’s “Contractualism and Social Risk” argues for the ex ante view with the following case:

Mass Vaccination. A virus will kill one million children unless we vaccinate them. Vaccine 1 saves every child but leaves each with a paralyzed leg. Vaccine 2 saves every child without paralysis, except that a random 0.1% of doses are defective.¹ Vaccine 3 is like Vaccine 2, except that we already know which children will die. (We have their names.) Suppose every child would rather risk a 0.1% chance of death than accept certain paralysis.

Intuitively, we are then required to choose Vaccine 2 over Vaccine 1, but there is very strong reason not to choose Vaccine 3: we would be asking 1000 children to bear a massive burden. Only the ex ante view, using epistemic rather than metaphysical probabilities, agrees: each child’s complaint against Vaccine 2 is only 0.1% as strong as a complaint against certain death, while Vaccine 3 leaves 1000 children with full-strength complaints. On the ex post view, Vaccines 2 and 3 equally generate 1000 full-strength complaints against death.

But ex ante contractualism would also have us sacrifice any number of statistical lives to rescue one single known victim. Frick illustrates this with the case of

Miners. Suppose we are in charge of n many miners, one of whom is now trapped. If we mount a rescue mission, we would exhaust the funding we would have used on upgrades that are projected to save 2% of the n lives.

According to ex ante contractualism, the single full-strength complaint of the trapped miner outweighs any number of 2%-strength complaints, no matter how large n is. So Frick proposes that ex ante contractualism be understood as an account of fairness rather than moral rightness. My goal is to examine accounts of fairness that would explain this proposal and evaluate whether ex ante contractualism meets the criteria given by that account.

I begin with John Broome’s account, on which fairness consists in satisfying claims in proportion to their strength. To Broome’s principles I add a Pareto condition: a procedure is unfair if it could satisfy some claim more without satisfying any other claim less. By Broome’s principle of Surrogate Satisfaction, a chance of receiving a good does not literally satisfy one’s claim to that good, but it is a surrogate for satisfaction; that is why a lottery can allocate a good fairly when it cannot be shared. The account also explains why natural lotteries, which allocate chances using pre-existing causal mechanisms, can be fair. Vaccine 2 is a fair natural lottery: every child has an equal claim to a healthy life. Vaccine 3 satisfies those equal claims unequally, and Vaccine 1 fails Pareto, so neither is fair. What ex ante contractualism requires, then, coincides precisely with what a Broomean account of fairness demands.

Against this view, Tim Henning argues that equal epistemic chances are not necessary for fairness, since a chance is not itself a good. Michael Otsuka, on the other hand, argues that equal epistemic chances are not sufficient, since equal metaphysical chances would be fairer still. I argue that they are both wrong: equal epistemic chances are necessary and sufficient for fairness when claims are equal. Even if we simply grant Henning that a chance is not a good, that does not entail we do not owe others chances, since we have interests in being given chances at goods. To use an example from Frick, if I secretly play Russian Roulette on you while you sleep but do not kill you, I do not make you worse off—but I wrong you. Against Otsuka, I argue that claims are satisfied or violated by agents acting under epistemic constraints, so the relevant chances are epistemic. This is not to say that metaphysical chances matter not at all, but they already contribute via epistemic probabilities, since epistemic probabilities constitutively track metaphysical chances. Decision-makers have an obligation to bring their epistemic probabilities reasonably close to metaphysical chances, and then to satisfy claims fairly using those epistemic probabilities.

Epistemic probabilities, however, give rise to Frege puzzles, as observed by Anna Mahtani, Johan Gustafsson, and Kacper Kowalczyk, the latter two of whom pose

Arthritis and Bronchitis. Ann and Beth are equally ill. You, the doctor, can split your one dose of a drug or give it all to the firstcomer, and a 1/2 chance of the full dose is much better than half a dose for sure. Smith arrived first, but you have forgotten which of Ann and Beth is Smith.

Under the names “Ann” and “Beth”, each has an even chance of the full dose, so giving it to the firstcomer is better for both—clearly we should do that. Under “Smith” and “Jones”, Jones is sure to get nothing—and this is unfair to Jones. The names we use to think of the two patients thus make the Pareto principle give inconsistent predictions.

The solutions on offer let any set of names decide, which collapses into utilitarianism; require every set to agree, which leaves too many moral dilemmas; or let context decide, so that learning that Jones has a sweaty ear could shift us to “Smith” and “Jones” and make it required to split the drug. But how could it be morally relevant that Jones has a sweaty ear? What all three miss is the contractualist tenet: that fairness consists in bilateral relations between one person and another, not between one person and their mental representations. A response is joint work in progress with Stephanie Van Fossen and Johann Frick.

Qualifying paper, written for Johann Frick; an earlier version was given at the Wollheim Society in March 2025.

Questions, objections, or a better example? Write to me: patrick.pan@patrickpan.com.