Negotiating AI Boundaries Through Participatory Futuring

Abstract

The increasing deployment of AI systems in consequential domains has intensified calls to democratize AI governance. However, existing participatory approaches often involve stakeholders only after key system boundaries have already been set, and tend to frame disagreement as a problem to be resolved rather than a condition to be engaged. In this paper, we argue that many AI governance challenges are fundamentally boundary-setting problems, characterized by irreducible value conflicts over what AI should do, under what conditions, and who gets to decide. Drawing on agonistic pluralism and participatory design, we propose to apply a four-stage Participatory Boundary Negotiation (PBN) approach to consequential AI deployments. We demonstrate PBN's application through an illustrative use case of AI-assisted grading in universities, showing how the method grounds deliberation in situated practices, surfaces conflicts through speculative futures, diagnoses their underlying conflict types, and supports negotiation through boundary objects. Rather than seeking consensus, this approach treats disagreement as a legitimate and productive feature of democratic AI governance, thereby contributing to participatory futuring research in consequential contexts.