We provide a general method for extending social preferences defined for riskless economic environments to the context of risk and uncertainty. We apply the method to the problems of managing unemployment allowances (in the context of macroeconomic fluctuations) and catastrophic risks (in the context of climate change). The method guarantees ex post fairness and pays attention to individuals' risk attitudes, while ensuring rationality properties for social preferences, revisiting basic ideas from Harsanyi's celebrated aggregation theorem (Harsanyi, 1955). The social preferences that we obtain do not always take the form of an expected utility criterion, but they always satisfy statewise dominance. When we require social preferences to be expected utilities, we obtain a variant of Harsanyi's result under a weak version of the Pareto principle, and a maximin criterion under a stronger Pareto requirement, whenever the ex post social ordering does not depend on people's risk attitudes. We also show how non-expected utility individual preferences can be accommodated in the approach.