Introducing the OECD AI Capability Indicators

The OECD AI Capability Indicators Report introduces a new framework for measuring artificial intelligence (AI) and robotic capabilities in direct comparison to human skills. Developed over five years with contributions from more than 50 experts in computer science, cognitive psychology, and related disciplines, the framework defines nine core dimensions of intelligence: language, social interaction, problem solving, creativity, metacognition and critical thinking, knowledge, learning and memory, vision, manipulation, and robotic intelligence, for evaluating whether an AI system exhibits narrow functionality or approaches a human-equivalent performance level.

What sets this framework apart is its use of a five-level developmental scale for each capability. Level 1 corresponds to very limited or basic functionality, while level 5 marks full human capacity. The intermediate levels go from narrow, specialised systems to more general and integrated intelligence. This approach reflects insights from human developmental psychology and cognitive science, enabling comparisons between AI systems and the range of skills demonstrated by people across different contexts and ages.

The indicators are designed not just to track current capabilities, but to anticipate future AI developments. By identifying which cognitive and physical skills AI systems are beginning to master and which remain out of reach.

Crucially, the OECD AI Capability Indicators Report provides a dynamic view of AI intelligence, recognising that progress is uneven across domains. This differentiated approach contrasts with earlier models that treat intelligence as a single measure and fails to account for the varied strengths and weaknesses of modern AI systems.