[Paper Review] Bridging the Global Divide in AI Regulation: A Proposal for a Contextual, Coherent, and Commensurable Framework

Sang-Chul Park|arXiv (Cornell University)|Mar 20, 2023

Law, AI, and Intellectual Property4 citations

TL;DR

This paper proposes a Contextual, Coherent, and Commensurable (3C) framework to bridge the global divide in AI regulation by categorizing AI systems into three task-based types—autonomous, discriminative (allocative, punitive, cognitive), and generative—based on their deployment context and human interaction. It ensures regulatory coherence through tailored objectives per category and promotes global interoperability via international risk measurement standards.

ABSTRACT

As debates on potential societal harm from artificial intelligence (AI) culminate in legislation and international norms, a global divide is emerging in both AI regulatory frameworks and international governance structures. In terms of local regulatory frameworks, the European Union (E.U.), Canada, and Brazil follow a horizontal or lateral approach that postulates the homogeneity of AI, seeks to identify common causes of harm, and demands uniform human interventions. In contrast, the United States (U.S.), the United Kingdom (U.K.), Israel, and Switzerland (and potentially China) have pursued a context-specific or modular approach, tailoring regulations to the specific use cases of AI systems. This paper argues for a context-specific approach to effectively address evolving risks in diverse mission-critical domains, while avoiding social costs associated with one-size-fits-all approaches. However, to enhance the systematicity and interoperability of international norms and accelerate global harmonization, this paper proposes an alternative contextual, coherent, and commensurable (3C) framework. To ensure contextuality, the framework (i) bifurcates the AI life cycle into two phases: learning and deployment for specific tasks, instead of defining foundation or general-purpose models; and (ii) categorizes these tasks based on their application and interaction with humans as follows: autonomous, discriminative (allocative, punitive, and cognitive), and generative AI. To ensure coherency, each category is assigned specific regulatory objectives replacing 2010s vintage AI ethics. To ensure commensurability, the framework promotes the adoption of international standards for measuring and mitigating risks.

Motivation & Objective

• Address the growing global divide in AI regulatory approaches between horizontal (EU, Canada, Brazil) and context-specific (US, UK, Israel, Switzerland, potentially China) models.
• Overcome the limitations of one-size-fits-all regulation by enabling context-sensitive, mission-critical risk management in diverse AI applications.
• Enhance international regulatory interoperability and accelerate harmonization through a standardized, yet adaptable, governance framework.
• Replace outdated 2010s AI ethics principles with outcome-oriented regulatory objectives aligned with real-world AI use cases.
• Establish a foundation for cross-jurisdictional alignment by promoting shared risk measurement standards across AI system categories.

Proposed method

• Bifurcate the AI life cycle into two distinct phases: pre-deployment learning and post-deployment task-specific deployment, avoiding reliance on the concept of foundation or general-purpose models.
• Classify AI systems into three regulatory categories based on application and human interaction: autonomous, discriminative (allocative, punitive, cognitive), and generative AI.
• Assign specific, outcome-driven regulatory objectives to each category to replace generic ethical principles from the 2010s era.
• Integrate international standards for risk assessment and mitigation to ensure commensurability across jurisdictions and regulatory systems.
• Structure the framework to support both national regulatory autonomy and cross-border regulatory coherence through shared classification and measurement protocols.
• Use the framework to guide legislation, enforcement, and international cooperation by aligning regulatory design with actual AI system behaviors and societal impacts.

Experimental results

Research questions

• RQ1 How can a regulatory framework reconcile context-specific AI governance with the need for global interoperability?
• RQ2 What criteria can effectively distinguish between different types of AI systems based on their deployment context and human interaction?
• RQ3 How can regulatory objectives be realigned from abstract ethics principles to concrete, measurable outcomes in specific AI applications?
• RQ4 To what extent can international standards enable commensurability across divergent national AI regulatory models?
• RQ5 Can a unified framework reduce regulatory fragmentation without undermining national policy autonomy in AI governance?

Key findings

• The 3C framework successfully redefines AI regulation by shifting from a one-size-fits-all model to a context-sensitive, category-based approach that aligns with real-world AI deployment patterns.
• Categorizing AI into autonomous, discriminative (allocative, punitive, cognitive), and generative types enables more precise and targeted regulatory responses.
• Replacing 2010s-era AI ethics principles with outcome-specific regulatory objectives enhances clarity, enforceability, and policy coherence across jurisdictions.
• The integration of international standards for risk measurement and mitigation significantly improves the potential for regulatory commensurability and cross-border alignment.
• The framework demonstrates feasibility in bridging divergent regulatory models—such as the EU’s horizontal approach and the US’s use-case-specific model—by providing a common analytical and normative foundation.
• The proposed life cycle bifurcation into learning and deployment phases avoids conceptual ambiguities around foundation models and supports more dynamic, risk-responsive regulation.