Delphi-based visual scenarios: An innovative use of generative adversarial networks

Abstract

In the Futures Studies context, the Delphi-based scenario (DBS) is a valuable method for setting future-oriented strategies and actions by gathering expert opinions in multiple iterative rounds, in order to make better decisions in the present. However, one of the main challenges is to find a suitable representation of the scenarios both for evaluation and decision-making. The development of future scenarios requires high creativity and simple textual descriptions could not be enough to grab the attention of the experts and decision-makers, due to a probable lack of communication and/or technical language. Since the textual descriptions concern the future, the human mind may have difficulty imagining events that have not already happened. With the spread of AI models, different implementations can support the DBS process, such as the use of Generative Adversarial Networks (GANs) to generate specific images starting from the narratives of the scenarios. In this paper, we propose a new method for the development of Delphi-based Visual Scenarios (DBVS) adopting GANs, with the aim to support the experts and decision-makers in planning and communication, in order to have immediate images of what the future could be like. We apply this method with regard to the future of Italian families.