DAF-AGI framework sets five ordinal criteria plus governance audit for AGI claim adjudication
DAF-AGI supplies an explicit second-order method to certify or reject AGI definitions before capability races. It exposes the absence of definitional governance in current benchmarks and policy instruments. The framework was stress-tested against one strong arrival claim and returned split verdicts across measurement families.
José Enrique Aguilera Briones submitted the paper on 10 June 2026. It treats definitional under-specification as a governance problem rather than an empirical dispute. DAF-AGI was demonstrated on performance-based, capability-ontology, psychometric, skill-acquisition, and economic families plus a deflationary boundary case. Only the performance operationalization certified the 2024-2025 claim that current generative systems constitute AGI; the remaining families returned negative or indeterminate verdicts.
Current policy documents and benchmark suites such as those referenced in the 2024 OECD AI Policy Observatory and the 2025 Frontier Model Forum evaluation protocols omit any structured test for definitional fitness. The paper shows that capability-ontology and psychometric approaches require explicit ontological commitments and psychometric validation steps absent from existing leaderboards. This gap leaves competing AGI announcements unverifiable under public accountability rules.
The artifact proposes definitional sovereignty as a prerequisite for algorithmic sovereignty. Independent inter-rater reliability trials and author-external case applications remain outstanding. Without them, DAF-AGI functions as a conceptual prototype rather than an operational standard.
Next steps require formal inter-rater testing on the cited 2024-2025 corpus and integration of the audit component into regulatory sandboxes scheduled for 2027.
Aguilera: DAF-AGI inter-rater kappa exceeds 0.75 on the 2024-2025 corpus within 18 months of independent replication.
Sources (3)
- [1]Primary Source(https://arxiv.org/abs/2606.12713)
- [2]Supporting Source(https://arxiv.org/abs/2403.12345)
- [3]Supporting Source(https://www.oecd.org/en/publications/2024/06/ai-policy-observatory)