The Organisation for Economic Co-operation and Development (OECD) has introduced a new framework designed to evaluate artificial intelligence (AI) capabilities in relation to human abilities, marking a significant step toward transparent, evidence-based AI policy development.
The OECD AI Capability Indicators, released in beta form, offer a structured set of nine indicators that assess AI performance across critical human domains: language, social interaction, problem solving, creativity, metacognition and critical thinking, knowledge, learning and memory, vision, manipulation, and robotic intelligence. Each domain is measured on a five-level scale, with level five representing full human equivalence.
The indicators are the result of a five-year effort by the OECD’s Artificial Intelligence and Future of Skills (AIFS) team, working in collaboration with more than 50 experts in AI and psychology. According to the OECD, the goal is to provide policymakers with a reliable tool to understand and track AI progress in ways that are both accessible to non-technical audiences and relevant to social, educational, and economic policy.
The report’s comparative table places current state-of-the-art AI systems—such as large language models and advanced robotics—between levels two and three on the five-point scale. While systems like GPT-4o demonstrate strong language capabilities, they fall short in areas such as analytical reasoning, metacognitive awareness, and dynamic learning. Social robots, meanwhile, remain limited in their ability to navigate nuanced social interactions or perform multi-step collaborative tasks.
The OECD’s framework is intended to fill a critical gap in the global AI landscape. Despite growing interest in monitoring AI development—evident in regulatory efforts like the European Union’s AI Act—no widely accepted, systematic tool currently exists to measure and compare AI capabilities in human terms. The OECD’s approach stands apart by grounding its indicators in human psychology and real-world functionality rather than benchmark performance alone.
The beta version invites feedback from AI researchers, psychologists, educators, and policymakers ahead of a planned 2026 update. Additional initiatives include developing new benchmark tests, launching an expert survey panel, and integrating dynamic updates through a dedicated online repository.
The OECD describes the indicators as a step toward defining and potentially measuring artificial general intelligence (AGI), providing policymakers a clearer basis to assess when AI systems might reach or exceed human-level performance across domains.
Need Help?
If you have questions or concerns about any global guidelines, regulations and laws, don’t hesitate to reach out to BABL AI. Their Audit Experts can offer valuable insight, and ensure you’re informed and compliant.
