[Paper Review] A Practical Multilevel Governance Framework for Autonomous and Intelligent Systems

Lukas Pöhler, Klaus Diepold+1 more|arXiv (Cornell University)|Apr 21, 2024

Collaboration in agile enterprisesBusiness, Management and Accounting3 citations

TL;DR

This paper proposes a practical multilevel governance framework for autonomous and intelligent systems (AIS) that maps actors across six decision-making levels—international, national, organizational, and others—enabling coordinated, adaptive governance through evolving tools and formal mechanisms. The framework supports agile, iterative oversight via regulatory sandboxes, periodic reviews, and bidirectional information flows, enhancing responsible AIS development and deployment across domains.

ABSTRACT

Autonomous and intelligent systems (AIS) facilitate a wide range of beneficial applications across a variety of different domains. However, technical characteristics such as unpredictability and lack of transparency, as well as potential unintended consequences, pose considerable challenges to the current governance infrastructure. Furthermore, the speed of development and deployment of applications outpaces the ability of existing governance institutions to put in place effective ethical-legal oversight. New approaches for agile, distributed and multilevel governance are needed. This work presents a practical framework for multilevel governance of AIS. The framework enables mapping actors onto six levels of decision-making including the international, national and organizational levels. Furthermore, it offers the ability to identify and evolve existing tools or create new tools for guiding the behavior of actors within the levels. Governance mechanisms enable actors to shape and enforce regulations and other tools, which when complemented with good practices contribute to effective and comprehensive governance.

Motivation & Objective

• Address the pacing problem in AIS governance, where rapid technological deployment outpaces slow-moving regulatory institutions.
• Overcome limitations of top-down, centralized governance by enabling distributed, multi-stakeholder participation across multiple levels of decision-making.
• Provide a structured yet flexible framework to identify relevant actors, develop governance tools, and ensure accountability across the AIS lifecycle.
• Integrate agile governance practices—such as regulatory sandboxes and iterative policymaking—into a scalable governance architecture for AIS.
• Facilitate bottom-up input and top-down coordination through formalized governance mechanisms that link levels and support oversight

Proposed method

• Map governance actors across six hierarchical levels: international, national, organizational, sectoral, community, and individual.
• Develop governance mechanisms that enable formal links between levels, including top-down enforcement and bottom-up proposal submission.
• Introduce iterative policymaking and periodic reviews to adapt governance tools in response to technological and societal changes.
• Incorporate regulatory sandboxes to test governance tools in controlled environments before large-scale deployment.
• Use good practices such as transparency, stakeholder engagement, and ethical standards to strengthen governance mechanisms.
• Leverage existing international bodies (e.g., ITU, UNODA) and forums (e.g., AI for Good Global Summit) to coordinate global input and avoid fragmentation

Experimental results

Research questions

• RQ1 How can governance be effectively decentralized and coordinated across multiple levels to address the rapid development of AIS?
• RQ2 What mechanisms enable both top-down oversight and bottom-up input in a multilevel AIS governance system?
• RQ3 How can agile governance practices like regulatory sandboxes and iterative reviews be integrated into a multilevel framework?
• RQ4 What role do existing international organizations and forums play in supporting a cohesive global governance ecosystem for AIS?
• RQ5 How can governance tools be evolved over time to remain effective amid technological and societal changes?

Key findings

• The proposed multilevel governance framework enables structured coordination among diverse actors across six decision-making levels, enhancing adaptability and inclusivity.
• Governance mechanisms support bidirectional flows—allowing lower-level actors to contribute proposals to higher levels and enabling higher levels to enforce or guide lower-level actions.
• The integration of agile governance practices such as regulatory sandboxes and periodic reviews enhances the responsiveness and evolution of governance tools.
• Existing international bodies like the ITU and UNODA can serve as effective platforms for coordinating global AIS governance, reducing fragmentation.
• The framework supports the creation of interoperable standards, declarations, and self-regulatory commitments through inclusive, multi-stakeholder forums.
• A governance coordination committee—such as the International Congress for the Governance of AI—can act as a neutral broker to mediate between diverse actors and promote consensus