Steps toward automating decision making

Standard models of decision-making provide normative methods for analyzing quantitative assessments of decision situations. These models are inadequate for most problems in artificial intelligence because they do not specify how to acquire assessments, and they depend on a complete, combinatorial model that requires too much data to assess. This paper presents a brief overview of some of these models and proposes an alternative methodology, constructive decision-making, that addresses some short-comings of normative models in knowledge-based paradigms. An implementation of the constructive decision-making approach is described.