My research examines Machine Pedagogical Intelligence: how educational AI can be structured, governed, inspected and evaluated while preserving human educational judgement. The work connects architectures and methods with implemented systems, including LearnAdapt, PedOS, CALLIOPE, Critsly and LearnAdapt Praxis.
This page brings papers and related public writing together by research theme. Repository note: items hosted on arXiv or SSRN are preprints or technical reports unless a peer-reviewed venue is explicitly named. Repository posting is not peer-review acceptance.
Application-linked research
CALLIOPE: source-grounded oral assessment
CALLIOPE: A Source-Grounded Oral Assessment System and Synthetic Readiness Evaluation — Technical report describing an inspectable response-to-review workflow with source grounding, recording and transcription, adaptive questioning, two-provider rubric scoring, educator review and exportable evidence. The reported evidence comes from synthetic fixtures and software checks; it does not establish scoring validity, learning gains or human-participant outcomes.
arXiv · SSRN · DOI · CALLIOPE application · Related co-design invitation
Critsly and StudioCrit: artefact-aware critique
Critsly and StudioCrit: An Artefact-Aware AI Critique Workspace and Simulation-Based Readiness Study for Design Education — Technical report on a critique-to-evidence workflow connecting artefacts, intentions, reflection, multi-perspective critique and action planning. StudioCrit adds a focused architecture-studio research mode. The evidence is simulation-based and does not demonstrate learning gains or human cognitive performance.
arXiv · SSRN · DOI · Critsly application · Related article
TeamLens in Critsly: voluntary disclosure and team reflection
TeamLens in Critsly: A Consent-Based Team-Composition Interface and Synthetic Readiness Evaluation for Design Collaboration — Technical report on an optional Critsly interface for voluntarily sharing a self-reported MBTI type with a particular board. It reports synthetic aggregation, access, lifecycle and bounded performance checks. It does not establish psychometric validity, human usability, learning gains or team-performance effects.
arXiv · SSRN · DOI · Related article
LearnAdapt Praxis: adult workplace learning
LearnAdapt Praxis is an early application for adults developing Design·AI capability through workplace challenges. Its technical report evaluates selected workflow, authority, recovery and provenance properties using synthetic project episodes; it does not establish learning effectiveness.
SSRN · DOI · Application overview · Explore Praxis
Pedagogical orchestration and governed educational AI
- From Untamed Black Box to Interpretable Pedagogical Orchestration: The Ensemble of Specialized LLMs Architecture for Adaptive Tutoring — accepted AIED 2026 full paper. arXiv · Published DOI · Related article
- From Tools to Teacher-Built Teammates: No-Code Pedagogical Plugin Authoring with LearnAdapt Agentic Studio and PedOS 1.1 Lumina — accepted AIED 2026 Interactive Events contribution. arXiv · DOI · Related article
- A Functional Taxonomy of Intelligent Systems in Education — preprint with Dorien Herremans. SSRN · DOI · Related article
Teacher co-design and accountable authoring
- Beyond Prompt-to-App: Accountable Translation in Teacher-Facing Agentic Authoring — preprint; not peer reviewed. arXiv · DOI
- From Misconceptions to Evidence: What Science Teachers Make Visible When Co-Designing Agentic Learning Apps — working paper; the related SISTC 2026 oral-presentation abstract was accepted, while the full manuscript has not been peer reviewed or accepted for proceedings. arXiv · DOI · Related article
Evaluation, assurance and uncertainty
- EduPluginBench: Executable Assurance for AI-Generated Educational Plugins — preprint; journal submission does not establish acceptance. arXiv · DOI · Related article
- Auditable Release Control for Pedagogical Leakage in LLM Tutors — preprint; not peer reviewed. arXiv · DOI · Related article
- RankCert: When Can Simulated Learners Safely Select an AI Tutor? Robust Decision Certification Under Structural Uncertainty — preprint. “Safe” is benchmark-scoped decision certification, not a general safety or human-learning claim. arXiv · DOI · Related article
Design critique and project-based learning
Critsly: An Artefact-Aware AI Critique Teammate for Design Education and Project-Based Learning — accepted AIED 2026 Interactive Events contribution. arXiv · DOI · Related article
Open technical resources
LearnAdapt Technical Blueprint v1.2 — citable public technical resource. Zenodo record · Version DOI · Concept DOI
Research profiles
Google Scholar · ORCID · ResearchGate · LearnAdapt Research · GitHub