Standards collaboration
AABoard reaches initial agreement to co-develop IEEE P3428 for AI-powered education
AABoard has reached an initial agreement with IEEE Standards Association and the IEEE Computer Society's Artificial Intelligence Standards Committee to collaborate on the development of IEEE P3428, Standard for Large Language Model Agents for AI-powered Education.
The collaboration brings AABoard's focus on AI literacy, educational assessment, and public evidence into a standards effort addressing how AI agents are designed, connected, deployed, and evaluated in education. AABoard's planned contribution centers on a practical question: what should an educational AI agent demonstrate before learners and institutions can rely on its educational claims?
A common foundation for educational AI agents
IEEE P3428 addresses agent architecture, interoperability, lifecycle management, foundation-model integration, and evaluation in educational settings. The project sits within the IEEE Computer Society's Artificial Intelligence Standards Committee and its LLM-4EDU working group. IEEE's project page lists P3428 as an active standards development project.
As educational agents take on more complex tasks, institutions need ways to examine their teaching behavior alongside their technical operation. An agent may complete a task successfully while leaving a learner dependent on its help. A useful evaluation must therefore consider both what the system accomplishes and what the learner is able to understand and do independently.
Three connected contributions: requirements, testing, and evidence
AABoard's priority contribution plan proposes three connected deliverables that turn educational expectations into testable requirements and reviewable evidence.
Educational evaluation and conformance requirements
The proposed requirements address pedagogical quality, learner agency, safety, accessibility, equity, and appropriate human oversight. They distinguish technical conformance, educational behavioral conformance, and demonstrated educational effectiveness, helping ensure that an educational claim is proportionate to the evidence supporting it.
Educational behavioral conformance tests
A focused initial suite of approximately 15 to 20 priority tests would examine realistic educational situations, including feedback, scaffolding, academic integrity, minor safety and privacy, and educator intervention. Each scenario would specify expected behavior, critical failures, scoring, and repeat testing to account for the variability of language-model outputs.
An educational agent evidence and assurance report
A common reporting template would document the agent and model version, intended learners and setting, evaluation results, limitations, corrective actions, and available pilot evidence. A claim-to-evidence matrix would make clear what the results support, while change-control records would identify when further evaluation is needed.
Bringing classroom experience into standards development
AABoard's case and pilot work provides a foundation for asking concrete questions about educational agents: Does assistance preserve learner reasoning? Can an educator interrupt or override the system? Does feedback help a learner recognize a misconception? Are safety and privacy practices appropriate for the learner's age and context?
Connecting these questions to common requirements and repeatable tests can help developers design more accountable systems and give educators, researchers, and procurement teams a clearer basis for comparing evidence. The contribution plan includes worked examples for a K-12 tutor and a higher-education writing assistant to keep the approach grounded in recognizable learning situations.
Moving the work forward
The initial agreement establishes a direction for collaboration. AABoard's proposed deliverables will be developed and refined through the working-group process as the P3428 project progresses. The aim is to connect AI innovation with educational accountability through clear requirements, repeatable testing, and transparent evidence.
Learn more about IEEE P3428, explore AABoard's standards work, or contact AABoard about contributing educational implementation experience to this effort.