Research Pillars
Our research sits at the interface of advanced AI methods and critical educational theory, examining how technology can serve, rather than subvert, local educational sovereignty.
Research Pillars
Epistemic Auditing of Educational LLMs
We systematically evaluate how LLMs respond to curriculum-aligned education queries (e.g., Ugandan secondary physics, East African history, local civic education).
Do they respect local curricula, or overwrite them with foreign narratives?
How do they handle contested histories, local political contexts, and indigenous knowledge?
We design metrics for epistemic authority, cultural alignment, and curricular fidelity, beyond generic correctness.
Graph-Based Reasoning for Learning Trajectories
We experiment with knowledge graphs + graph neural networks + symbolic reasoning to model learning pathways.
- Predict how misconceptions propagate across subjects (e.g., algebra misunderstandings undermining data science).
- Suggest personalized, curriculum-consistent remedial paths.
- Simulate how changes in curriculum structure might affect cohorts over time.
Multilingual & Multimodal Educational AI
We develop and evaluate speech, text, and sign-language interfaces that serve diverse linguistic communities.
- Code-switching conversational tutors that mirror real classroom language practices.
- Local language question-answering systems grounded in national content.
- Experiments in multimodal models that integrate text, audio, diagrams, and gestures for STEM learning.
AI, Assessment, and Academic Integrity
We investigate how generative models interact with essay writing, problem-solving, and project-based assessments.
- Design "AI-aware" assessment rubrics that measure conceptual depth, critical thinking, and originality.
- Study how students actually use AI (as a calculator, co-writer, explainer, or shortcut).
- Develop policy and tooling to align institutional assessment practices with the realities of AI.
Futures, Speculation, and Educational Governance
We use speculative design, scenario building, and futures workshops with teachers, policymakers, and students to imagine post-digital, post-platform education futures.
- Imagine education futures that center local control and epistemic justice.
- Anticipate risks (hyper-surveillance, data scoring, dependency on foreign platforms).
- Prototype alternative infrastructures, such as locally owned learning clouds and community-controlled data trusts.
Methodological Approach
Quantitative ML Experiments
Benchmarks, ablations, robustness tests, and comparative evaluations across models and contexts.
Qualitative & Ethnographic Work
In-depth classroom observations, teacher interviews, and student focus groups to understand real-world contexts.
Participatory Design
Teachers and learners as co-researchers in iterative design cycles, ensuring tools meet real classroom needs.
Policy & Legal Analysis
Examining data protection frameworks, intellectual property rights, and AI governance structures.
Research Impact
Our research generates both technical and conceptual contributions that shape policy, practice, and future inquiry.
Each research project generates both technical artifacts (code, models, datasets) and conceptual artifacts (frameworks, guidelines, policy briefs)—all contributing to a growing body of evidence for locally-grounded education AI.
