A methodology for the optimization of the shortterm emergency response in the event of a nuclear accident is presented. The method seeks an optimum combination of protective actions in the presence of a multitude of conflicting objectives and under uncertainty. Conflicting objectives arise in the attempt to minimize simultaneously the potential adverse effects of an accident and the associated socioeconomic impacts. Additional conflicting objectives arise whenever an emergency plan tends to decrease a particular health effect, such as acute deaths, while it increases another, such as latent deaths. The uncertainty is due to the multitude of possible accident scenarios and their respective probability of occurrence, the stochastic variability in the weather conditions, and the variability and/or lack of knowledge of the parameters of the risk assessment models. A multiobjective optimization approach is adopted. An emergency protection plan consists of defining a protective action (e.g., evacuation and sheltering) at each spatial cell around the plant. Three criteria (evaluators) are used as the objective functions of the problem, namely, acute fatalities, latent effects, and socioeconomic cost. The optimization procedure defines the “efficient frontier,” i.e., all emergency plans that are not dominated by another in all three criteria. No value trade-offs are necessary up to this point. The most preferred emergency plan is then chosen among the set of efficient plans. Finally, the methodology is integrated into a computerized decision support system, and its use is demonstrated in a realistic application.