Developing optimal release policies of multi-purposes reservoirs is very complex, especially for reservoirs with explicit stochastic environment, (e.g. uncertainty in future inflows). Most of optimization techniques that are applied to search for optimal releases, use summative measures of reservoir performance, e.g. expected benefits or costs. These techniques have limited aptitude to represent risks associated with deciding in a release policy. In fact, the approach of these techniques may be at odds with the true attitudes of the decision-makers toward the risk aspects of release policies. The risk aspect of the decisions affects the design and operation of multi-purpose reservoirs. A mean for a more complete representation and evaluation of the possible consequences associated with a release decision is therefore necessary. A decision-making model capable of replicating the manner in which risks associated with reservoir release decisions are perceived, interpreted and compared by a decision-maker is proposed. The model is based on Neural Network (NN) theory, and enables a more complete representation of the risk function present in a particular reservoir release decision.