Human-in-the-loop AI for multilingual education.
How can generative AI scale educational content without removing expert judgement? ACCLAIMED is a University of Malta AI research project that puts expert review at the centre of an AI content pipeline.
- The problem
- How to scale educational content with generative AI without removing expert judgement.
- What the project does
- Experts supply knowledge, AI helps author the material, and a review layer checks it before it is delivered.
- My contribution
- Project foundations · research and technical documentation · UI/UX · workflow design · selected development
- Team
- University of Malta research team
- Status
- Applied AI research project.
- Related
- I am a co-author on a systematic review of hyper-realistic avatars in education, listed under Research.
The project
ACCLAIMED stands for Artificial Intelligence Content Cloning of Language-Agnostic Media for Education Democratisation. It is a University of Malta research project designed around multilingual educational delivery.
The project documentation describes three core components.
Course Generator
Educators and subject-matter experts provide foundational knowledge and pedagogical input. AI authoring tools then support generation of course content, curricula, assessment instruments and educational activities.
Guardrail
A human-expert review and safety layer for inspecting and refining AI-generated content: accuracy, ethical integrity, societal expectations, linguistic fluency and human oversight.
AI Renderer
A component designed to transform generated material into human-like audio-visual learning content, with supporting educational resources.
Multilingual delivery
The system is designed around delivering the same educational material across languages.
How the pieces connect
Generated content is stopped at the Guardrail until a human expert reviews it, then continues to the AI Renderer and multilingual delivery.
- Subject-matter expertProvides knowledge and pedagogical input.
- Course GeneratorAI authoring tools produce learning material.
- Guardrail: expert validationHuman experts review and refine the generated content.
- Validated contentMaterial that has passed expert review.
- AI Renderer, multilingual deliveryValidated material is transformed into audio-visual learning content in multiple languages.
Conceptual flow based on the project model, recreated for this page.
My contribution
My work on ACCLAIMED spanned the foundations of the project, research and technical documentation, UI/UX workflow design and selected development work.
The design challenge was translating a complex AI research concept into a system that educators and subject-matter experts could interact with, without needing to understand the AI behind it.
My contribution
Project foundations, research and technical documentation, UI/UX and workflow design, and some software development.
The ACCLAIMED system
The wider University of Malta research project and its Course Generator, Guardrail and AI Renderer components, developed by the project team.
Why it matters for business AI
AI systems that generate consequential outputs need validation, expert review and clear points of human control. ACCLAIMED made that review layer a first-class component. The same principle runs through TDi and the business workflows I build: the domains differ, the design principle carries over.