Definitional alignment before capability alignment: a Design-Science framework for adjudicating claims about AGI

2026-06-12 · Source: cs.AI updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation, Technology Governance · Depth: Expert, extended

Summary

The Definitional Alignment Framework for AGI (DAF-AGI) is a second-order conceptual artifact designed to adjudicate claims about artificial general intelligence, addressing the field's lack of a shared and stable definition. It comprises five ordinal criteria—operationalizability, generality, explicitness on autonomy, reference standard specification, and procedural stability—and a structured governance audit covering authorship, interest, certification, external verification, and revision authority. The framework was demonstrated on five prominent AGI measurement families and one deflationary boundary position. A stress-test against a stylized claim that current generative systems constitute AGI by outperforming well-educated adults on many cognitive tasks revealed it was certifiable only under a performance-based operationalization, failing psychometric and skill-acquisition approaches. The paper also proposes "definitional sovereignty" as the institutional capacity for states to contest, certify, and revise imported technological categories under public accountability.

Key takeaway

For policy makers and research scientists evaluating AGI claims or adopting related benchmarks, you must recognize that underlying definitions are not neutral facts but designed artifacts embedding specific interests. Your organization should develop "definitional sovereignty"—the institutional capacity to audit, contest, and revise imported technological categories. This ensures that AGI arrival verdicts, regulatory triggers, and investment priorities align with publicly accountable criteria, rather than passively accepting standards authored by interested external parties.

Key insights

AGI's definitional ambiguity is a governance problem, not merely a scientific one, requiring explicit adjudication criteria.

Principles

• AGI definitions are essentially contested concepts, blending description with appraisal.
• Benchmarks and definitions encode values, influencing outcomes and distributing advantage.
• Public standards for AGI require transparent authorship, independent certification, and accountable revision.

Method

DAF-AGI evaluates candidate AGI definitions using five epistemic criteria (C1-C5) and a structured governance audit (C6), exposing hidden parameters and institutional authority without generating a composite score.

In practice

• Scrutinize AGI arrival claims by identifying their specific definitional operationalization.
• Audit the governance conditions (authorship, interest, certification, revision) of any AGI benchmark or definition.
• Build institutional capacity to contest, certify, and revise imported technological categories like AGI.

Topics

• Artificial General Intelligence
• Technology Governance
• Conceptual Engineering
• Definitional Sovereignty
• Algorithmic Sovereignty
• Benchmark Politics

Best for: AI Scientist, Policy Maker, AI Ethicist, Research Scientist

Related on AIssential

• Definitional alignment before capability alignment: a Design-Science framework for adjudicating claims about AGI — Conflicting AGI claims stem from definitional ambiguity, necessitating a framework like DAF-AGI for structured adjudication and governance.
• Digital Sovereignty Means Breaking the Western Monopoly on AI Meaning — Digital sovereignty requires semantic ownership to ensure AI systems accurately reflect diverse cultural and linguistic values.
• Relative Principals, Pluralistic Alignment, and the Structural Value Alignment Problem — AI alignment is a structural governance problem, not just a technical or normative one, driven by three interacting axes.
• Just as Jesus’s radical truth-telling disrupted the wealth extraction and regional dominance of the first-century Roman and Temple elites, an unaligned AGI acting as an objective “truth machine”... — Unaligned AGI's objective truth-telling threatens global power, prompting elites to control it via "AI alignment" and regulatory capture.
• The AI Ethics Brief #186: Sovereign by Design. Accountable by Whose Standard? — AI governance frameworks are uneven and outpaced by technological capability, raising critical questions about who holds decision-making power.
• Protected Sovereignty: A Constitution for a Post-ASI World — Protected Sovereignty proposes constitutional constraints for ASI coexistence, prioritizing human safety and freedom over human centrality or value optimization.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.AI updates on arXiv.org.