From Clicks to Consensus: Collective Consent Assemblies for Data Governance

TL;DR

This work argues that traditional notice-and-consent is inadequate for modern data practices due to interlinked data and communal harms. It introduces collective consent, operationalized through consent assemblies that leverage deliberative mini-publics and speculative design, supported by future backcasting to outline feasibility and steps. Two vignettes illustrate practical applications: replacing notices for UX data and governing GenAI training data, highlighting both potential and challenges. The paper outlines a roadmap for legal, regulatory, and societal transformations needed to enable collective consent and proposes future empirical work to validate and refine the approach.

Abstract

Obtaining meaningful and informed consent from users is essential for ensuring they maintain autonomy and control over their data. Notice and consent, the standard for collecting consent online, has been criticized. While other individualized solutions have been proposed, this paper argues that a collective approach to consent is worth exploring for several reasons. First, the data of different users is often interlinked, and individual data governance decisions may impact others. Second, harms resulting from data processing are often communal in nature. Finally, having every individual sufficiently informed about data collection practices to ensure truly informed consent has proven impractical. We propose collective consent, operationalized through consent assemblies, as one alternative framework. We establish the theoretical foundations of collective consent and employ speculative design to envision how consent assemblies could function by leveraging deliberative mini-publics. We present two vignettes: i) replacing notice and consent, and ii) collecting consent for GenAI model training, to demonstrate its wide application. Our paper employs future backcasting to identify the requirements for realizing collective consent and explores its potential applications in contexts where individual consent is infeasible.

Figures (1)

• Figure 1: The proposed steps of a consent assembly. The solid lines indicate steps that could happen in every consent assembly, and the dotted lines indicate steps that might potentially happen depending on whether the situation requires it or not.