Seeing economic development like a large language model. A methodological approach to the exploration of geographical imaginaries in generative AI

Abstract

The recent hype surrounding the disruptive potential of AI technologies in the form of large language models or text to image generators also raises questions for geographical research and practice. These questions include the power relations and inequalities inscribed in these systems, their significance for work and labor relations, their ecological and economic impact, but also the geographical and spatial imaginaries they reproduce. This article focuses on the latter and formulates a series of theoretical and methodological considerations for dealing with the output of these systems. As we assume that outputs generated by large language models will play an increasing role in the future, both in public and media discourses as well as in the discourses and practices of spatial planning and economic policy making, we consider it important to gain a critical understanding of these socio-technical systems. The empirical object of investigation of this paper is generated output that deals with questions of regional development and economic challenges in three European regions that are currently particularly affected by the transition to a climate-neutral economy and are designated by the European Union as Just Transition Fund Territories. We are particularly interested in how geographical imaginaries about these regions are formulated, how economic and social problems of these regions are presented and how this is translated into planning advice and development plans.