Openvalue: a Participatory Governance Infrastructure for Ai Value Formation
Year: 2026
Type: preprint
Abstract: A central challenge in participatory artificial intelligence (AI) governance is translating input fromdiverse participants into decisions that meaningfully influence AI systems. Existing research has explored various participation mechanisms, but participants’ influence remains constrained by insti-tutional arrangements and technical conditions. We introduce OpenValue, a participatory governance infrastructure for AI value formation. OpenValue supports the construction of dynamic scenarios in which models experience the conse-quences of their own decisions and reflect on their actions and outcomes, providing an entry point for human governance. The platform combines scenario-based interaction, structured process records, issue identification in model reflections, domain-specific routing, and independent review. Draw-ing on the model’s behavior and its consequences, participants can examine the value orientations revealed in its reflections, alongside the underlying connections and potential contradictions amongits understanding, judgments, actions, and outcomes. Reviewers’ assessments and their rationalesare linked to the original records, providing auditable governance materials for resolving disputes,tracing responsibility, and training value modules in subsequent work.A financial scenario demonstration illustrates the complete workflow from model experience and reflection to human participation in governance. Functional tests further verified the technical feasi-bility of the platform’s core mechanisms and the traceability of its governance processes. OpenValue provides scalable sociotechnical infrastructure for studying how diverse participants shape AI valuesand for exploring partial decoupling between the governance of AI value formation and establishedmodel development systems.
Authors: Chen, Lusi, Tong, Mu, Zhang, Zirui, Xiao, Yexuan
Institutions: South China University of Technology, Tsinghua University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7233608968
Cites:
Cited by:
Related to:
