New tools for understanding AI and learning outcomes

The education sector is still early in understanding how AI tools like ChatGPT affect long-term learning outcomes, as current research often relies on narrow performance signals like test scores. To address this gap, OpenAI, in collaboration with the University of Tartu and Stanford's SCALE Initiative, developed the Learning Outcomes Measurement Suite. This framework supports longitudinal measurement across different educational contexts by tracking signals like model behavior, learner response, and measurable cognitive outcomes over time, including autonomous motivation, persistence, and metacognition.

Initial research using a study mode feature in ChatGPT with college students showed mixed results: meaningful gains in microeconomics scores but not in neuroscience, highlighting the limitations of short-term assessments. The new Measurement Suite aims to capture the evolving, personalized interactions between learners and AI, providing a standard framework for educators and researchers to evaluate AI's influence against defined pedagogical standards.

The suite is currently undergoing extensive validation through large-scale randomized controlled trials, such as one involving nearly 20,000 students in Estonia. OpenAI intends to release the measurement suite as a public resource, fostering a deeper, thoughtful integration of AI that cultivates higher-order thinking and supports diverse learning goals across global education systems.

(Source：OpenAI)