Veil-of-Ignorance Reasoning Favors the Greater Good

Karen Huang, Joshua D. Greene, and Max Bazerman test whether considering equal chances of occupying each affected position changes subsequent moral judgments.^[1]

A two-stage procedure

Across seven experiments (reported total N = 6,261; four preregistered), participants first considered what they would want if they had an equal chance of occupying each affected person’s position. They then answered a conventional version of the dilemma as a decision-maker. The comparison was between these later responses and responses from control groups, rather than between the self-interested choice behind the veil and an ordinary moral judgment.^[1]

The equal-chance assumption follows Harsanyi’s formulation: probabilities are specified, rather than unknown as in Rawls’ original version. In the footbridge case, five of six possible positions benefit from pushing one person to stop a trolley. In study 1, 38% of the veil group judged pushing morally acceptable, compared with 24% of the control group. Thus the intervention shifted judgments without making the sacrificial choice a majority preference in that experiment.^[1]

Policy, donations and competing explanations

Study 2 applied the procedure to a hospital oxygen-allocation dilemma and a law requiring autonomous vehicles to minimize deaths. In the vehicle scenario, swerving saves nine pedestrians but kills the passenger. Approval of the policy was 83% after the veil exercise versus 58% in the control condition. This is a stated judgment about a specified regulation, not evidence about buying a vehicle or its real-world crash decisions.^[1]

Study 3 introduced a real donation decision: one randomly selected participant’s choice determined where a $200 donation went. The described alternatives funded sight-saving treatment for two people in India or one person in the United States. After imagining equal chances of being each potential beneficiary, 63% chose the more effective charity, compared with 54% of controls (n = 833). Not every participant made an individually funded donation.^[1]

Studies 4–6 tested alternatives involving consistency with a prior answer, numerical/probabilistic reasoning and adopting a utilitarian person’s perspective. In study 5, reversing the probabilities gave the participant a nine-in-ten chance of being the single passenger; the impartial veil condition still elicited more utilitarian policy judgments (73% versus 64%). The authors interpret the comparisons as support for impartial reasoning, while acknowledging that component features such as probabilistic thinking may contribute.^[1]

Where the effect stops

Study 7 tested transfer: participants completed hospital and charity veil exercises before judging the vehicle policy. Its binary outcome did not differ significantly from either control group (57% versus 53% and 54%); a scale-rating difference appeared against the simple control, but not the anchoring control. The findings therefore support a procedure applied to a particular choice more clearly than a general, transferable change in moral outlook.^[1]

Veil of Ignorance provides a possible value-selection procedure for Social Alignment, but these experiments do not establish that the resulting judgments are morally correct. Their dilemmas do not distinguish utilitarianism from Rawlsian maximin, and the paper does not claim the procedure always favors utilitarian choices. The later Using the Veil of Ignorance to align AI systems with principles of justice tests a different choice—how an AI assistant distributes help—rather than training a system on these moral-dilemma responses. Engineering a social contract: Rawlsian distributive justice through algorithmic game theory and artificial intelligence addresses the broader institutional proposal.^[1]