My role
Needs analyst, designer, developer, teacher, evaluator
Learners
10 pupils, grades 1 to 9
Context
City of Riihimäki, Finland · 2 sessions a week · Aug 2025 to now
Tools
Tutor LMS, H5P, WordPress, Claude
Approach
ADDIE · flipped · game-based · mastery
1 · The challenge
Nine grade levels in one room
Beginners and fluent speakers from grade 1 to grade 9 learn together, with uneven skills and only two sessions a week. Families want their children to keep the language and a Vietnamese identity while growing up in Finland.
2 · Needs analysis
Start with the learners
I analysed each learner’s language gaps, learning preferences and cultural background. The findings set the course structure, the levels and the flipped, game-based model.
See the User Needs Analysis →3 · Design decisions
What I chose, and why
Flipped classroom
Short online lessons at home; class time goes to speaking, discussion and one-to-one help.
Why: With two sessions a week, contact time is too valuable for content delivery.
Mastery with retries
Retries with instant feedback on every H5P activity. At 100%, the pupil adds a star to the class board.
Why: Mastery learning and visible progress build competence and motivation (self-determination theory).
Game-based mechanics
Points, stars, levels and challenge-style tasks.
Why: Game-based learning makes home practice something children choose to do.
A personal package for each learner
Each pupil practises listening, speaking, reading and writing at their own level.
Why: A grade 1 beginner and a grade 9 fluent speaker cannot share one path.
GenAI as co-designer, teacher as editor
I co-create activities with Claude, then check every item for accuracy, age-appropriateness and cultural fit. No pupil data goes into the AI.
Why: My own Ethical AI principles for designing learning materials.
4 · Iteration
Three redesigns, driven by what learners did
-
1Version 1 → 2
Skill tracks
What I changed: I organised the course into four skill tracks (listening, speaking, reading, writing), each with levels 1 to 4.
Why: Pupils’ skills were uneven: a child could be level 3 in listening but level 1 in writing, so separate tracks fit no one.
-
2Version 2 → 3
Personal packages
What I changed: I replaced the tracks with a personal practice package for each pupil, mixing all four skills at that pupil’s level.
Why: Every pupil now practises all four skills each week, at a level that fits them.
-
3Ongoing
Clearer tasks for older pupils
What I changed: The discussion task was too easy for grade 9. I added a structured discussion guide and stated exactly what product they must create.
Why: Older pupils’ discussions became more focused and produced work that could be assessed.
5 · Evidence so far
What the learners did
pupils reached 100% on every activity, retrying until they mastered it, and marked each result with a star on the class board.
Parents report that their children enjoy the Vietnamese lessons and retry an activity until they earn the star, without being asked.
Honest limitation
The free Tutor LMS plan does not store attempt data, so this evidence comes from the star board, my observation and parent feedback.
6 · Next iteration
From observation to data
- Log attempts and time on task (H5P xAPI to a learning record store, or an LMS plan that keeps results).
- A short four-skill check at the start and end of each term.
- A five-question pupil and parent survey each term.
- Use the data to simplify, extend or replace activities.
7 · What I learned
Lessons I take to every project
- Personalise by the learner, not by the content structure.
- Visible mastery (a star, a board) can motivate more than a grade.
- GenAI speeds up production; the designer stays responsible for quality and ethics.
- Decide what to measure before launch, not after.
Want to talk?
Hiring for instructional design or learning engineering? I’d be glad to walk you through my work.