JÖNKÖPING UNIVERSITY • MASTER’S THESIS • 2025–2026
ROLE
UX Designer & Full-Stack Developer
TIMELINE
Nov 2025 – Ongoing
METHODS
DSR, Mixed Methods, System Logs, Interviews
TOOLS
React, Supabase, OpenAI API, Figma
OVERVIEW
Students begin every semester motivated. By week four they're reactive, chasing deadlines instead of planning ahead. Existing LMS platforms like Canvas, Blackboard, and Moodle are built to track content delivery, not how students are actually studying. They show what was submitted, not how someone got there. Regula is an AI-enhanced, gamified learning platform built to change that, developed and studied as part of a master's thesis at Jönköping University, co-designed with Ngo Nguyen Tuan Kiet.
THE PROBLEM
Self-Regulated Learning research identifies three critical phases: planning, monitoring, and reflection, where student performance tends to diverge. Students who lack scaffolding at the right moment slide from planning into passive execution, and from execution into missed deadlines, by which point it's already too late. No existing LMS nudges students to plan, reflect, and adjust on a weekly cycle. Most tools only alert on deadlines already missed.
"She doesn't need more reminders. She needs smarter, better-timed ones."
THE JOURNEY
Nov 2025
STARS Hackathon, Finland (remote)
Early prototype tested with 46 participants over 48 hours. Validated the core concept and identified the planning scaffold as the highest-priority feature.
Nov 2025 – Apr 2026
Full platform built
Full React + Supabase + OpenAI API codebase developed from scratch. Two experimental conditions implemented on a single codebase: AI-Enhanced and Controlled.
Apr 2026
Two-week intervention study
27 enrolled participants across two conditions. Mixed methods: system logs, reflection entries, pre/post questionnaires, semi-structured interviews.
2026
Science Park: pitch competition
Presented Regula as a commercialization concept. Won Best Pitch Award.
Graduation 2026
Best Business Potential Award, Jönköping University
University-level recognition following the thesis defence. Regula cited for innovation in applying AI to structured learning support.
Ongoing
Second study in preparation
Platform in active development. Methodology refined from first study learnings, evaluating commercial path following Science Park recognition.
THE PLATFORM
The platform runs two experimental conditions on a single shared codebase. Both groups get the same gamification layer (XP, badges, streaks, and a leaderboard), held constant on purpose to isolate the effect of AI on learning behavior. The AI-Enhanced group gets three additional features: an AI-generated weekly roadmap, feedback on assignments, and personalized resource recommendations. The Controlled group sets its own goals manually, with no AI features at all.
Regula is still in active development toward commercialization, so this case study shows a simplified structure and selected screens rather than a full product walkthrough.
AI-ENHANCED
AI Roadmap: day-indexed task breakdown generated weekly
AI Feedback: automated response to submitted assignments
AI Resource Recommendations: contextual to roadmap tasks
XP, badges, leaderboard, and streaks (shared with Controlled)
CONTROLLED
Manual Goals page: students set their own tasks without AI input
Same gamification layer: XP, badges, leaderboard, streaks
No AI features: no roadmap, no automated feedback, no recommendations
SELECTED SCREENS
Home Dashboard
Course Modules
Daily Reflection Pop-up
Gamification Element: Streak count
RESEARCH & FINDINGS
The study used a mixed methods design: system logs, reflection entries, pre/post questionnaires, and semi-structured interviews. 27 participants enrolled across two conditions; 11 remained consistently active. Statistical analysis used Mann-Whitney U tests given non-normally distributed data. Qualitative coding of interview transcripts identified recurring themes across both conditions.
6.05 / 4.50
Autonomy score — AI-enhanced vs controlled
Leaderboard
identified as most behaviorally consequential element across both conditions
Week over week, the AI-Enhanced group's task completion rate improved from Week 1 to Week 2, while the Controlled group's declined. That suggests AI scaffolding may sustain engagement over time in a way manual goal-setting doesn't.
"Participants didn't delegate planning to the AI. They used its output as a starting point, then overlaid their own judgment."
HONEST FINDING
The planning-execution gap
AI-assisted task decomposition significantly increased how many tasks students created, but that didn’t consistently translate into tasks completed. The platform supported planning well; it didn’t yet address the monitoring step that turns intentions into finished work.
Despite less consistent follow-through, AI-Enhanced participants self-reported higher metacognitive self-regulation than the Controlled group, a gap between what people felt and what they actually did that's worth sitting with.
DESIGN DECISIONS
01
Weekly cycles not daily
Sustainable rhythm over short sprints. Daily cycles create pressure; weekly cycles create space for reflection. The AI roadmap generates weekly task breakdowns, not daily checklists, so students can reason about their week without feeling surveilled.
02
AI as scaffold, not replacement
Students keep strategic authority over their own planning. The AI generates structure they can edit, reject, or rebuild, not instructions they have to follow. Participants used the roadmap as a starting point, not a script. That was the intended behavior.
03
Gamification held constant across conditions
This isolates the AI's effect clearly. If gamification varied between conditions too, the findings would be uninterpretable, since there'd be no way to separate the effect of AI from the effect of XP and leaderboards. Holding it constant was the only methodologically sound choice.
04
Leaderboard kept despite debate
Qualitative data showed the leaderboard was the most motivating element for most participants. Removing it for methodological purity would have made the platform less useful, and participants would have noticed its absence. Handled transparently, competition became a feature, not a flaw.
WHAT’S NEXT
A second study is in preparation with a refined methodology. Two changes come directly from this study's findings: giving reflection prompts visible trigger conditions for when reflection is actually expected, and adding a live monitoring layer that flags two consecutive missed days and suggests a way back on track, closing the planning-execution gap this study surfaced. The platform stays fully functional and in active development, not a prototype or a concept, and is being evaluated for commercial potential following the Science Park award.
