📄ArXiv AI•Stalecollected in 21h
A Design-Science Framework for Adjudicating AGI Claims
💡Stop guessing if a model is AGI. Use this new framework to objectively validate capability claims.
⚡ 30-Second TL;DR
What Changed
Introduces DAF-AGI, a conceptual artifact for evaluating AGI definitions.
Why It Matters
This framework provides a rigorous methodology for researchers and policymakers to standardize how we evaluate AGI, preventing misleading hype in the industry.
What To Do Next
Apply the DAF-AGI criteria to your internal model evaluation benchmarks to ensure your AGI claims are robust against independent scrutiny.
Who should care:Researchers & Academics
Key Points
- •Introduces DAF-AGI, a conceptual artifact for evaluating AGI definitions.
- •Uses a structured governance audit to assess authorship, certification, and verification.
- •Tests current 'strong arrival' claims against five distinct measurement families.
- •Argues that definitional sovereignty is essential for algorithmic accountability.
🧠 Deep Insight
Web-grounded analysis with 12 cited sources.
🔑 Enhanced Key Takeaways
- •DAF-AGI aims to resolve the ambiguity in AGI definitions, which currently hinders effective AI governance and design, by providing criteria to test if a given definition holds up under scrutiny.
- •The framework's structured governance audit specifically examines the authorship, interests, certification, external verification, and revision authority of AGI claims, providing a blueprint for holding these definitions accountable.
- •DAF-AGI introduces 'definitional sovereignty' as a subset of 'algorithmic sovereignty,' advocating for institutions to have the power to contest and revise technological definitions under public scrutiny to ensure transparency and accountability in AGI governance.
- •When tested against various measurement families, including psychometric and skill-acquisition approaches, DAF-AGI revealed that current 'strong arrival' claims for generative systems might only be supported by performance-based metrics, with other approaches disagreeing or refusing binary adjudication.
📊 Competitor Analysis▸ Show
| Feature / Framework | DAF-AGI (ArXiv AI, 2026) | Google DeepMind's Levels of AGI (Morris et al., 2023/2024) | Cognitive Framework (Center for AI Safety et al., 2025) |
|---|---|---|---|
| Primary Focus | Evaluating AGI definitions and their associated governance structures. | Classifying AGI based on performance depth and generality breadth. | Quantifying AGI progress by matching human cognitive versatility and proficiency. |
| Methodology | Structured governance audit (authorship, certification, verification, revision authority) and five criteria for assessing definitions. Tests claims against measurement families (e.g., psychometric, skill-acquisition, capability-ontology, economic). | Defines five performance levels (Emerging, Competent, Expert, Virtuoso, Superhuman) across a spectrum of generality. | Grounds methodology in Cattell-Horn-Carroll (CHC) theory, dissects intelligence into 10 core cognitive domains, and adapts human psychometric tests to generate an 'AGI Score'. |
| Key Output | Adjudication of AGI claims based on definitional rigor and governance accountability. | A common language to compare models, assess risks, and measure progress along the path to AGI. | A quantifiable 'cognitive profile' and 'AGI Score' (0-100%) for AI systems, highlighting strengths and weaknesses. |
| Current Application | Conceptual framework for resolving definitional disputes and ensuring accountability in AGI claims. | Used to categorize current systems (e.g., LLMs like ChatGPT or LLaMA 2 as emerging AGI). | Estimates GPT-4 at 27% AGI score and anticipated GPT-5 at 58%, identifying deficits in areas like long-term memory. |
🛠️ Technical Deep Dive
- Structured Governance Audit Components: The DAF-AGI framework incorporates a governance audit that examines the authorship, interests, certification, external verification, and revision authority related to AGI claims.
- Criteria for Assessing Definitions: The framework includes five criteria specifically designed for assessing the robustness and clarity of AGI definitions.
- Measurement Families for Claim Adjudication: DAF-AGI tests AGI claims against distinct measurement families, which include psychometric approaches, skill-acquisition approaches, capability-ontology, and economic metrics.
- Focus on Definitional Rigor: Unlike frameworks primarily focused on technical capabilities, DAF-AGI emphasizes the importance of clear and accountable definitions as a prerequisite for effective AGI governance.
🔮 Future ImplicationsAI analysis grounded in cited sources
Increased scrutiny and standardization of AGI claims.
DAF-AGI's framework provides a structured, objective method for evaluating AGI definitions and claims, which could lead to more rigorous verification processes by institutions and regulators.
Enhanced algorithmic accountability and governance in AI development.
By emphasizing definitional sovereignty and structured governance audits, the framework could drive greater transparency and responsibility in how AI systems, particularly AGI, are developed and deployed.
A shift in focus from purely performance-based AGI metrics to a more holistic evaluation encompassing ethical and societal implications.
The framework's emphasis on governance, authorship, and definitional sovereignty suggests a move towards considering who defines AGI and how those definitions impact society, rather than just technical capabilities.
⏳ Timeline
1997
Mark Gubrud uses the term 'artificial general intelligence' (AGI) in a discussion of automated military production.
2000
Marcus Hutter proposes AIXI, a mathematical formalism of AGI, defining intelligence as an agent's ability to achieve goals in a wide range of environments.
2007
The term AGI is popularized by Ben Goertzel and Cassio Pennachin with the publication of their co-edited book 'Artificial General Intelligence,' which also provided a definition.
2023
Google DeepMind researchers propose a framework for classifying AGI into five performance levels (Emerging, Competent, Expert, Virtuoso, Superhuman) based on depth and breadth of capabilities.
2025-10
Researchers from the Center for AI Safety, Université de Montréal, Stanford, and MIT propose a 'Cognitive Framework' for defining AGI, grounded in Cattell-Horn-Carroll theory, using 10 cognitive domains.
2026-06
The DAF-AGI framework is proposed to address the ambiguity in AGI definitions and evaluation standards through structured governance audits and ordinal standards.
📎 Sources (12)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
📰
Weekly AI Recap
Read this week's curated digest of top AI events →
👉Related Updates
AI-curated news aggregator. All content rights belong to original publishers.
Original source: ArXiv AI ↗