Jean Marie Ntsama
AI Writing Lab
Text-to-Image Pedagogy in Priority Education Zones
The AI Writing Lab is a pedagogical innovation developed within the IA_4_ZEP initiative. It uses generative AI as a motivation and feedback tool for descriptive writing in large, low-resource classrooms (60+ learners).
1. The Pedagogical Problem
Teaching descriptive writing in ZEPs faces three obstacles:
- Large classes make individual feedback impossible.
- Learners struggle to understand what makes a description “precise.”
- Abstract writing exercises feel disconnected from learners’ lived experience.
2. The Innovation: Write to See
The method is simple: if the AI can generate the image, the description was clear. If it cannot, the description needs revision.
The workflow:
- Corpus: Learners study a literary character (e.g., Faydé from Cœur du Sahel, a prescribed Cameroonian text).
- Reading: They identify physical traits, clothing, posture, expressions, and context.
- Writing: Each learner produces a short descriptive text.
- Generation: The teacher feeds each text into an AI image generator.
- Confrontation: The generated image is compared with the learner’s intention.
- Revision: Learners refine their descriptions based on what the AI misunderstood.
3. Classroom Evidence
Step 1 — The Corpus Character: Faydé (15 years old)
Two images generated from the original literary text:
Step 2 — Learner Productions (3rd year, LTBNM)
Five examples generated from learner-written texts:
Each image demonstrates a learner’s ability to produce a structured, precise, culturally-anchored descriptive text.
4. Why This Works: Cognitive Foundations
This method operationalizes three concepts from the Cognitive Trapeze Model:
- Short Loop (Learner–AI–Knowledge): The AI generates a visual response to the learner’s text — an immediate, tangible feedback loop.
- Embodied Knowledge: The abstract description becomes a concrete, visual, affective experience.
- Pedagogical Curability: Learners with language difficulties are supported by the visual outcome, which makes their progress visible.
5. Scalability & Open Access
This methodology is:
- Replicable across all grade levels and literary works.
- Adaptable to bilingual contexts (French / English).
- Low-cost — it requires only a teacher’s smartphone or tablet.
All methodology, lesson plans, and anonymized learner productions will be released on Zenodo under a CC-BY license.
Published Datasets:
- AI-Generated Comic “Tendani Ito” (2026): https://doi.org/10.5281/zenodo.23065165
- ASR Reading Corpus (Texts 1–18) (2026): https://doi.org/10.5281/zenodo.22922869
6. Zenodo Deposits & Institutional Compliance
Two complementary datasets from the AI Writing Lab have been published on Zenodo (October 2026):
Deposit 1 — Text-to-Image Pedagogy
Ntsama, J. M., & IA_4_ZEP Initiative. (2026). Cognitive Trapeze — AI Writing Lab: Text-to-Image Pedagogy and Learner Productions in Cameroonian ZEP [Dataset]. Zenodo. https://doi.org/10.5281/zenodo.23065607
Deposit 2 — Narrative Descriptive Productions
Ntsama, J. M., & IA_4_ZEP Initiative. (2026). Cognitive Trapeze — AI Writing Lab: Narrative Descriptive Productions and AI-Generated Visualizations in Cameroonian ZEP [Dataset]. Zenodo. https://doi.org/10.5281/zenodo.23230681
Compliance & Ethics:
- Data collection conducted under MINESEC Authorization Ref. N° 658/2026/AR/H/DRES (Regional Delegation of Adamaoua).
- Written parental consent and child assent obtained for all participants.
- All learner names replaced with pseudonyms; no biometric or identifiable data included.
- License: Creative Commons Attribution 4.0 International (CC-BY 4.0).
Contents (Deposit 2):
- 6 AI-generated images from learner narrative-descriptive texts
- Anonymized learner productions with prompts
-
README with methodology and FAIR documentation