What scenario-building characteristics should be used in GenAI prompting?

Abstract

This article discusses the prompting process when using GenAI, especially large language models (LLM) in scenario building. It revolves around the different types of scenarios and their characteristics, between AI-generated scenarios and AI-assisted scenarios, e.g. in Expert knowledge modelling and in Econometrics, and the necessity to have a structured multistep prompting procedure in case one wants to get good quality, well-targeted scenarios. These steps naturally depend on the chosen methodology and the outcome that is sought, yet there are some principles that help set up good prompting. One of these principles is choosing the right scenario-building characteristics and the correct order of these when prompting. For this purpose, this article provides a three-step typology for specifying the scenario types.