A methodological approach to obtain bounds on the value of information based on an inexact representation of the decision maker's utility function is presented. Stochastic dominance procedures are used to derive the bounds. These bounds provide more information than the single point estimates associated with traditional decision analysis approach to valuing information, in that classes of utility functions can be considered instead of one specific utility function. Empirical results for valuing seasonal climate forecasts illustrate that the type of management strategy given by the decision maker's prior knowledge interacts with the decision maker's risk preferences to determine the bounds.