Democratic Friction in Speech Governance by AI

Abstract

Artificial Intelligence (AI) systems are increasingly deployed to govern the digital public sphere. Their robust and speedy application contributes to a frictionless flow of information - a signature trait of the digital economy, and of social media in particular. For democracy, however, friction is a key feature. It is reflected in principles such as separation of powers and discursive contestation, aiming to ensure self-governance and civil liberties. This chapter contends that a critical space for self-governance and social deliberation of social norms is ”lost in translation”, as we shift to speech governance by AI. Driven by probabilistic decision-making processes and recursive optimization dynamics, AI systems fail to facilitate a shared ground for discursively voicing dissent and addressing it. Accordingly, this chapter argues that public policy should focus on promoting democratic friction. To enable social dialogue on speech norms and facilitate contestation, AI systems should be reconfigured in a way that embeds democratic legal principles in their design.

Keywords: AI, artificial intelligence, ML, machine learning, governance, public sphere, free speech, content moderation, contestation, democracy, friction