On automating candle lighting analysis: insight from search with genetic algorithms and approximate models

Decision makers often need to find solutions to problems that they understand only approximately. For these difficult situations, we consider a model-based approach for a type of analysis which we term ´candle lighting analysis´, which suggests actions we might take to create new and better options. This paper shows how a space of actions can be searched by a genetic algorithm, automatically providing trade-off and sensitivity information to the decision maker. We use this approach in a DSS and show how a decision maker can obtain insight into the problem itself and knowledge about effective actions to take.<>