ABSTRACTProblem definitions are often used as part of the problem structuring process in decision sciences. We report an attempt to validate a multi‐attribute model of problem definition quality in the context of an experiment based upon the definition of the ‘alcohol problem’ on a campus. Three independent raters assessed values for the attributes of the model for 57 written problem definitions, and their rating was compared to three subjective ‘expert’ ratings made by individuals with some responsibility for alcohol policy. Our study shows that the experts exhibited considerable stakeholder bias in their assessments of the definitions, suggesting that a definition that is evaluated as good may not be subjectively acceptable to others. We present some evidence that the best definitions are those that provide direction with regard to solution, neither suggesting firm solutions to the problem nor being too open‐ended. It is shown that the multi‐attribute model is a reasonable tool for eliminating poorer definitions. Further, we relate our work to recent conceptual developments in understanding probl