ROUTER
Hub Router
Identifies the educational task, directs it to the appropriate specialist, and maintains a coherent workflow.
AI LAB
Designing teaching agents as a governed educational system.
AIP–TDA is a teacher-led, multi-agent design architecture for AI-supported project and problem-based learning.
The lab does not treat an AI agent as a single all-purpose tutor. Instead, it separates course knowledge, pedagogical methods, design production, routing, and governance into coordinated roles. This makes responsibilities visible and creates explicit review points for teachers.
01
AIP–TDA · Multi-agent architecture
COURSE
AI–MFT Course Agent
Connects disciplinary concepts, cases, materials, and learning evidence within a specific course.
METHOD
PBL / PjBL Method Agent
Supports problem framing, project structure, inquiry sequences, milestones, and reflection.
DESIGN
Teacher Design Pack Agent
Produces reviewable lesson structures, task briefs, scaffolds, rubrics, and teacher-facing drafts.
GOVERNANCE
AI–PBL Governance & Boundary Agent
Checks educational fit, role boundaries, evidence quality, risk, and teacher review requirements.
02
Governance decisions
Every output is a draft. The governance layer uses four legible decisions rather than an opaque score.
PASS
Proceed
The draft meets the defined educational and safety criteria.
CONDITIONAL PASS
Proceed with explicit conditions
The draft is usable after named revisions, checks, or teacher decisions.
SAFE-REWRITE
Reframe before use
The intent is valid, but the task or output requires a safer educational formulation.
BLOCK
Do not proceed
The request conflicts with role boundaries, evidence standards, or responsible-use rules.
Design principles and selected cases are public; internal prompts, knowledge bases, workflow files, credentials, and unreleased testing materials remain private.