•  
  •  
 

Document Type

Article

Abstract

The online elaboration of speech norms is enduring a decisive transformation, threatening the vital prospects of democratic contestation, which enable democracies to thrive. In this Article, we demonstrate how a critical space for social deliberation and negotiation of the desirable boundaries of free speech is "lost in translation" as we shift from governance by law to governance by Artificial Intelligence (Al). The configuration of AI speech filtering systems facilitates a frictionless flow of information-a signature trait of the digital economy, and of social media in particular. It is driven by a probabilistic decisionmaking process based on formal definitions and optimization dynamics, which are designed to enable speedy detection of harmful content. AI speech moderation systems effectively formulate data-driven decision rules, which reflect a single, pre-defined and potentially biased tradeoff. It currently lacks, however, adequate contesting mechanisms and fails to facilitate the vital normative space necessary for deliberating the disagreements in society regarding the scope of free speech. In contrast, governance of online speech by law is discursive, permitting different tradeoffs to coexist. Speech governance by law further facilitates a shared ground for voicing dissent and addressing it. By its institutional design, and various procedures and practices, governance by law in liberal democracies facilitates democratic contestation, and it is therefore better equipped to sustain divided societies in the absence of deeper normative consensus. The absence of democratic contestation in speech governance by AI undermines the legitimacy of speech norms, precludes public engagement in checking and testing which values are embedded in algorithmic tradeoffs, and interferes with the pluralistic aspiration to develop social norms through democratic processes of public engagement and deliberation. This Article proposes to introduce speech contestation by design in order to legitimize the way AI systems currently shape online speech norms. Inspired by the contestation mechanisms of the law, such as separation of powers and adversarial legal procedures, this Article suggests separation of functions and contesting algorithms as exemplary design features of Al systems of speech governance. Embedding such design features into AI systems of speech moderation may enable ongoing social dialogue between diversified views regarding the limits of free speech. Legal policy pertaining to automated speech moderation by digital platforms should therefore focus on promoting such design interventions.

Download

Share

COinS