The future of education is AI-powered — and that’s a good thing

1. Emphasize Humans-in-the-Loop. We reject the notion of AI as replacing teachers. Teachers and other people must be “in the loop” whenever AI is applied in order to notice patterns and automate educational processes. We call upon all constituents to adopt Humans-in-the-Loop as a key criteria.

2. Align AI Models to a Shared Vision for Education. We call upon educational decision makers, researchers, and evaluators to determine the quality of an educational technology based not only on outcomes, but also based on the degree to which the models at the heart of the AI tools and systems align to a shared vision for teaching and learning. 

3. Design AI Using Modern Learning Principles. Further, achieving effective systems requires more than processing “big data”— it requires more than data science. Applications of AI must be based on established, modern learning principles, the wisdom of educational practitioners, and should leverage the expertise in the educational assessment community around detecting bias and improving fairness.

4. Prioritize Strengthening Trust. Technology can help us to achieve our educational objectives only when we trust it. And our listening sessions revealed the ways in which distrust of artificial intelligence is natural. Because trust develops as people meet and relate to each other, we call for a focus on building trust and establishing criteria for trustworthiness of emerging educational technologies within the associations, convenings, and professional organizations that bring educators, innovators, researchers, and policymakers together.

5. Inform and Involve Educators. Going beyond receiving notice and explanation of the use of AI, educational leaders must prioritize informing and involving educational constituents so they are prepared to investigate how and when AI fits specific teaching and learning needs, and what risks may rise. Addressing data privacy is important, but strengthening trust requires more than data privacy: educators need AI systems that can be inspected, explained, and guide how humans can override recommendations generated using AI. The Office of Educational Technology can play a leading role in providing information and stimulating involvement, in concert with activities organized at all levels of the educational system.

6. Focus R&D on Addressing Context and Enhancing Trust and Safety. Research that focuses on how AI-enabled systems can adapt to context (diversity among learners, variability in instructional approaches, differences in educational settings) is essential to answering the question “Do specific applications of AI work in education, and if so, for whom and under what conditions?” We call upon researchers and their funders to prioritize investigations of how AI can address the long tail of learning variability and to seek advances in how AI can incorporate contextual considerations when detecting patterns and recommending options to students and teachers. Further, researchers should accelerate their attention to how to enhance trust and safety in AI-enabled systems for education.

7. Develop Education-specific Guidelines and Guardrails. Data privacy regulation already covers educational technology; further, data security is already a priority of school educational technology leaders. Modifications and enhancements to the status quo will be required to address the new capabilities alongside the risks of AI. We call for involvement of all perspectives in the ecosystem to define a set of guidelines (such as voluntary disclosures and technology procurement checklists) and guardrails (such as enhancements to existing regulations or additional requirements) so that we can achieve safe and effective AI for education