Project Description
The Adaptive AI (ADAPT-AI) project focuses on the design and evaluation of an AI-empowered assessment system to support three-dimensional science learning aligned with the Next Generation Science Standards (NGSS). The project responds to a persistent challenge in K–12 science education: teachers, particularly in under-resourced and rural contexts, often lack the time, tools, and support needed to design high-quality formative assessments that capture students’ scientific reasoning in use.
ADAPT-AI develops an adaptive, conversational AI system that supports teachers in co-designing, refining, and interpreting science assessment tasks. Rather than automating assessment decisions, the system is designed to work with teachers, allowing them to guide assessment design while drawing on AI support for alignment, adaptation, and feedback. The project brings together system design, learning sciences research, and independent evaluation to understand how AI can meaningfully support assessment practice in real classroom settings.
Additional information about the ADAPT-AI project is available on the AIR Communications website: Can AI Help Teachers Design Formative Assessments in Science Education? Evaluating the ADAPT-AI Tool
Project Information
Project Period: 2025–2027
Project Lead: Tingting Li
Funding Agencies:
This project is jointly supported by Microsoft AI for Good Lab, the National Academy of Education, and the American Institutes for Research.
Project Vision
Teachers play a central role in assessment design, yet often lack sufficient time, tools, and support to develop high-quality formative assessments that capture students’ scientific reasoning in use, particularly in under-resourced and rural contexts.
Rather than automating assessment decisions, ADAPT-AI is designed to position AI as a collaborative partner, supporting teachers’ professional judgment while enhancing transparency, adaptability, and usability in assessment practice.
Project Goals
- Goal 1: System Design and Development (Microsoft AI for Good)
Microsoft AI for Good supports the design and development of the adaptive AI assessment system. This component focuses on building a scalable, AI-powered, multi-agent system that enables teachers to design NGSS-aligned science assessments that are responsive to classroom context, student diversity, and instructional goals. - Goal 2: Human–AI Collaboration Research (National Academy of Education)
The NAEd component investigates the mechanisms of human–AI collaboration in assessment design. This work examines how teachers interact with AI systems, how professional judgment and AI suggestions are negotiated, and how trust, agency, and decision-making evolve when AI is positioned as a collaborative partner rather than an automated tool. - Goal 3: System Evaluation (American Institutes for Research)
AIR leads the independent evaluation of the ADAPT-AI system. This includes examining usability, implementation, and impacts on teacher practice and student assessment experiences. The evaluation focuses on how the adaptive AI system functions in authentic classroom settings and provides evidence to inform future scaling and refinement.
Team & Collaborators
Project Team:
Peng He, Zeyuan Wang, Julia Lee, Freeman Chen
Collaborators:
Microsoft AI for Good Lab;
National Academy of Education;
American Institutes for Research
Partners
Washington Educational Service Districts (ESD 101, 112, and 123)