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]

Method and contribution

Seven experiments examine sacrificial dilemmas, autonomous-vehicle policy, and charitable choices. The paper reports shifts toward options benefiting more people and tests competing explanations through preregistered controls. Its central evidence concerns the effects of a deliberation procedure, not AI training.^[1]

Alignment relevance and limits

Veil of Ignorance provides a possible value-selection procedure for Social Alignment. The experiments do not show that the shifted judgments are morally correct, that the procedure always promotes utilitarian choices, or that Rawlsian maximin is refuted. The tested dilemmas do not distinguish those theories. Attitudes toward vehicle regulation do not establish purchasing behavior.

Compare the later Using the Veil of Ignorance to align AI systems with principles of justice and conceptual Engineering a social contract: Rawlsian distributive justice through algorithmic game theory and artificial intelligence.