🌐 OpenLab Agentic Education

Preserving human agency in AI-supported learning.

AI assists. Humans govern. Evidence speaks. Communities validate.

Not robot-first. Trust-first.

What is this?

OpenLab Agentic Education is a research and prototype framework exploring how learners, educators, families, communities, institutions, and AI agents can co-create and examine learning evidence without surrendering human judgment or learner agency to automation.

It is the public home of the OpenLab / S.A.T. — Skills, Agency & Trust research line on:

• 📜 Proof-of-Learning — inspectable evidence of learning centered on the learner
• 🤝 Human-governed AI — AI may assist Evidence Review; Human Validation remains final
• 🧭 Learner agency — learners can explain, question, challenge, and help govern their evidence record
• 🎖 Recognition — badges or credentials follow validated evidence; they are not the evidence itself
• 🛡 Trust-first governance — privacy, provenance, consent, accessibility, uncertainty, and accountability by design
• ✨ MIRA — Meaningful Intelligence for Reflection and Agency, a proposed learner-facing companion that helps learners reflect and act without becoming their evaluator

Anchor paper: From Co-Intelligence to Proof-of-Learning

Vocabulary: docs/TERMINOLOGY.md defines the project's canonical language.

The learning trust loop

Mission → Learning Claim → Evidence → Reflection / Iteration
                         ↓
                AI-assisted Evidence Review
                         ↓
                HUMAN VALIDATION GATE
                         ↓
                    Learning Proof
                         ↓
          Recognition → Portfolio → Next Pathway

The goal is not to make learning more machine-readable. The goal is to make learning more human-visible without surrendering human agency.

How the repo is organized

The project moves through three connected layers, with human governance running through all of them:

flowchart LR
    subgraph Research["🔬 Research"]
        A[papers/]
        B[proposals/]
        C[docs/]
    end
    subgraph Framework["🧭 Framework"]
        D[framework/]
        E[governance/]
        F[missions/]
    end
    subgraph Prototype["⚙️ Prototype"]
        G[agents/]
        H[badges/]
        I[prototypes/]
        J[data/]
    end

    Research --> Framework --> Prototype
    Prototype -. human validation .-> Framework
    Framework -. human validation .-> Research

Loading

Layer	What lives here	Folders
🔬 Research	Concept papers, proposals, references	papers/ · proposals/ · docs/
🧭 Framework	Rubrics, learning loops, governance, ethics	framework/ · governance/ · missions/
⚙️ Prototype	Agent specs, recognition schemas, synthetic evidence review	agents/ · badges/ · prototypes/ · data/

gh600/ bridges an external agentic-AI certification track to the OpenLab framework.

Start here:

1. VISION.md — why this exists
2. MANIFESTO.md — principles we will not compromise
3. docs/TERMINOLOGY.md — the language of the system
4. framework/proof-of-learning-rubric.md — PoLR v0.1
5. agents/mira.md — MIRA prototype specification
6. ROADMAP.md — where this is going
7. papers/ — research grounding

Quick start

git clone https://github.com/porroto/Agentic-Education.git
cd Agentic-Education

• Educators → framework/
• Researchers → papers/ and proposals/
• Builders → agents/ and prototypes/ — synthetic data only

Safety & Ethics Boundary

• ✅ Synthetic and de-identified examples only — no real student data belongs in this repo
• ✅ Human Validation is final — AI reviews, questions, organizes, and suggests; people validate
• ✅ Minimum necessary evidence — more learner data is not automatically better evidence
• ✅ Classroom pilots require applicable school policy compliance, consent/permission processes, and ethics review
• ✅ Recognition, not speculation — badges/credentials represent validated learning evidence and are not financial instruments
• ❌ No financialized learning tokens for children
• ❌ No hidden behavioral, biometric, emotional, or psychological profiling
• ❌ No AI system gets to declare the human's learning complete

Who is this for?

• Learners who deserve agency over how their learning is represented
• Teachers designing evidence-based, project-driven classrooms
• Researchers studying human-AI collaboration and learning
• Builders who believe student agency is non-negotiable
• Communities & families who belong in the trust conversation

Roadmap (high level)

• Concept paper: From Co-Intelligence to Proof-of-Learning
• Three-layer architecture (Research → Framework → Prototype)
• Canonical terminology v0.1 — Skills, Agency & Trust; Learning Proof; Trust Envelope
• MIRA prototype specification
• PoLR v0.1 review draft
• Evidence Review Agent v0 synthetic adversarial test
• Trust Envelope schema v0.1
• Classroom-safe pilot kit
• Community validation protocol
• Public research brief + call for critique/collaborators

See ROADMAP.md for details.

Contributing

Contributions, critiques, and classroom perspectives are welcome — especially perspectives that expose where the framework could reproduce grading, surveillance, inequity, or false certainty. Please read CONTRIBUTING.md and our CODE_OF_CONDUCT.md.

Citing this work

See CITATION.cff.

License

MIT — see LICENSE.

---

Built by Roger Vargas · STEM educator & founder, S.A.T. Labs / OpenLab
Let's Learn. Lead. Make. 🌎