Definitional alignment before capability alignment: a Design-Science framework for adjudicating claims about AGI
Researchers present DAF-AGI, a governance framework for defining artificial general intelligence, arguing that competing definitions of AGI produce contradictory verdicts on the same systems. The framework tests whether current generative AI systems qualify as AGI and finds certification only under performance-based metrics, while other approaches reject the claim, highlighting the necessity of definitional clarity before capability assessment.
The article addresses a fundamental problem in AI governance: the absence of a shared definition for AGI creates confusion in both academic discourse and policy discussions. When claims about AGI's arrival rely on different operational definitions, stakeholders reach incompatible conclusions about identical systems. This definitional ambiguity has practical consequences for investment decisions, regulatory frameworks, and public expectations about AI capabilities.
The researchers respond by proposing DAF-AGI, a two-component framework that combines ordinal criteria for evaluating definitional fitness with a governance audit structure examining authorship, interests, verification, and revision processes. Their analysis of five measurement families reveals that current generative systems like GPT-4 meet AGI criteria only under narrow performance-based operationalizations, while capability-ontology, psychometric, and skill-acquisition approaches reject AGI classification. This divergence underscores how definition selection determines outcomes rather than evidence alone.
For the AI industry and investors, this work signals that the AGI debate requires institutional rather than purely technical resolution. The framework's emphasis on "definitional sovereignty" suggests that organizations and nations should establish transparent processes for accepting, contesting, or modifying AI capability categories rather than adopting imported definitions uncritically. This has governance implications: regulators cannot effectively oversee AGI risks without consensus on what AGI actually is.
Moving forward, the critical path involves independent application of DAF-AGI across different research teams, inter-rater reliability testing, and extension to emerging systems. Success would establish shared language for policy discussions, but the framework requires external validation before becoming authoritative in governance contexts.
- →Current generative AI systems only qualify as AGI under performance-based definitions; other measurement approaches reject this classification
- →DAF-AGI proposes a structured governance process for definitional evaluation including authorship transparency and revision authority
- →Definitional ambiguity about AGI creates contradictory policy and investment conclusions based on identical evidence
- →Definitional sovereignty enables institutions to contest and revise technological categories rather than passively accepting external definitions
- →The framework requires independent testing and inter-rater validation before applications in regulatory or institutional contexts