Is Artificial Intelligence threatening democracy?

Abstract

In a democracy, human beings make decisions with the aim of serving the will of the people and promoting the collective welfare of society. While machines can learn from data and generate potential democratic solutions, they fall short in grasping the intricacies of the subjective reality of democratic politics. Entrusting Artificial Intelligence (AI) systems with decision-making carries the risk of following optimal solutions shaped by falsified objective realities that AI algorithms aim to optimise. Even if the data were comprehensive and sufficient, modelling approaches struggle to fully encapsulate the complexities of subjective realities within global democracies and societies. In essence, leaving democratic politics to be governed by ostensibly logical AI classifiers is a significant gamble. In the second high-level policy dialogue that took place on the 22 and 23 of May 2023 in Florence under the auspices of the STG Chair in Artificial Intelligence and Democracy, scholars and policymakers discussed and shared their ideas to map multiple available solutions for how democratic politics can live with an AI-powered world and, more than that, how AI can turn to a beneficial tool for democracy. Most of the participants agree that AI can be formed and transformed to a useful tool for democracies. In this policy brief, we summarise the key ideas that emerged from the discussions in this high-level policy dialogue.