Performance-based design in earthquake engineering is a structural optimization problem that has, as the objective, the determination of design parameters for the minimization of total costs, while at the same time satisfying minimum reliability levels for the specifi ed performance criteria. Total costs include those for construction and structural damage repairs, those associated with non-structural components and the social costs of economic losses, injuries and fatalities. This paper presents a general framework to approach this problem, using a numerical optimization strategy and incorporating the use of neural networks for the evaluation of dynamic responses and the reliability levels achieved for a given set of design parameters. The strategy is applied to an example of a three-story offi ce building. The results show the importance of considering the social costs, and the optimum failure probabilities when minimum reliability constraints are not taken into account. Performance-based design in earthquake engineering is a structural optimization problem that has, as the objective, the determination of design parameters for the minimization of total costs, while at the same time satisfying minimum reliability levels for the specifi ed performance criteria. Total costs include those for construction and structural damage repairs, those associated with non-structural components and the social costs of economic losses, injuries and fatalities. This paper presents a general framework to approach this problem, using a numerical optimization strategy and incorporating the use of neural networks for the evaluation of dynamic responses and the reliability levels achieved for a given set of design parameters. The strategy is applied to an example of a three-story offi ce building. The results show the importance of considering the social costs, and the optimum failure probabilities when minimum reliability constraints are not taken into account.