Artificial Utopia: Simulation and Intelligent Agents for a Democratised Future
Yannick Oswald
TL;DR
The paper tackles how to study utopian, bottom-up democratic systems—specifically citizen assemblies and democratic firms—using computational simulations and AI to safely explore ideas in-silico. It advocates the Artificial Utopia research agenda, drawing on Computational Game Theory, Agent-Based Modelling, Reinforcement Learning, Large Language Models, and System Dynamics to model cognitive, deliberative, and macro-dynamics aspects of these institutions. By mapping simulation approaches to concrete challenges and outlining open questions and risks, the work aims to bridge social sciences with computational methods, enabling theory–empiricism synergy and informing policy and organizational design. The ultimate goal is to guide human societies toward democratic innovation while mitigating real-world risks and ethical concerns.
Abstract
Prevailing top-down systems in politics and economics struggle to keep pace with the pressing challenges of the 21st century, such as climate change, social inequality and conflict. Bottom-up democratisation and participatory approaches in politics and economics are increasingly seen as promising alternatives to confront and overcome these issues, often with utopian overtones, as proponents believe they may dramatically reshape political, social and ecological futures for the better and in contrast to contemporary authoritarian tendencies across various countries. Institutional specifics and the associated collective human behavior or culture remains little understood and debated, however. In this article, I propose a novel research agenda focusing on utopian democratisation efforts with formal and computational methods as well as with artificial intelligence - I call this agenda Artificial Utopia. Artificial Utopias provide safe testing grounds for new political ideas and economic policies in-silico with reduced risk of negative consequences as compared to testing ideas in real-world contexts. An increasing number of advanced simulation and intelligence methods, that aim at representing human cognition and collective decision-making in more realistic ways, could benefit this process. This includes agent-based modelling, reinforcement learning, large language models and more. I clarify what some of these simulation approaches can contribute to the study of Artificial Utopias with the help of two institutional examples: the citizen assembly and the democratic firm.