Responsible AI for Measurement and Learning: Principles and Practices

Abstract

The rapid proliferation of artificial intelligence (AI) in educational measurement presents both transformative opportunities and complex ethical challenges. This paper articulates foundational principles for the responsible integration of AI in measurement and learning, drawing on established guidelines set forth by leading organizations such as NIST, OECD, UNESCO, the U.S. Department of Education, and others. We propose a principled framework encompassing fairness and bias mitigation, privacy and security, transparency, explainability, accountability, educational impact and integrity, and continuous improvement. Through the synthesis of current research, best practices, and cross-sector standards, we highlight practical measures to ensure that AI-driven assessment systems are equitable, valid, and reliable. Special emphasis is placed on the significance of representative data, ongoing bias analysis, secure-by-design development, and stakeholder involvement throughout the AI lifecycle. This approach is designed to foster trust, uphold educational values, and safeguard individual rights. By emphasizing ethical and sustainable practices, we advocate for a vision of AI as a driver of human development—supporting learners, educators, and society at large in the pursuit of educational and economic mobility. The principles and recommendations outlined here offer guidance not only for our organization but serve as a resource for the broader educational and measurement community, charting a course for responsible AI innovation that advances both the science and the practice of measurement in support of lifelong learning.

Suggested citation: Johnson, M. S. (2025). Responsible AI for measurement and learning: Principles and practices (Research Report No. RR-25-03). ETS.