Polyglot VR — orange wax seal logo on parchment with medieval castle illustration

♠ Applied AI

Polyglot VR

Polyglot VR is a choice-based virtual reality game designed to reimagine second-language learning. The project addresses the lack of engaging, practical speaking experience in standard curricula by allowing users to roleplay and converse with AI-driven NPCs. The outcome is an interactive space that develops speaking and listening skills through contextual immersion.

Role
Interaction Designer
Timeline
15 weeks
Team
Solo Project (Thesis)
Year
2022

The Problem & The User

The Problem

Traditional language learning relies heavily on book learning, which often fails to engage or provide practical speaking experience. As we get older, finding the time to commit to learning a language is difficult, and the fear of embarrassment can hinder practice.

The User

The target users are intermediate language learners (A2-B2 CEFR) who own VR headsets and enjoy gaming, typically aged 16-34. They already possess foundational knowledge but struggle with natural cadence, conversational skills, and contextual vocabulary.

64%

Of language learners experience high "speaking anxiety"

70%

Of adults regret losing their foreign language skills

<10%

Language students achieve true proficiency

The Approach & Process

Research & Discovery

Research began with a literature review on language pedagogy and a market analysis of existing tools like Duolingo and Mondly VR. I conducted 5 user interviews to understand learner motivations, discovering a strong desire for contextual immersion and a fear of embarrassment when speaking.

Competitive analysis — testing Duolingo, Mondly VR, and AR language learning apps on various devices

Competitive analysis — testing Duolingo, Mondly VR, and AR language apps to identify gaps in immersive learning

The Solution

The result was a fantasy-themed virtual reality game where players navigate a branching quest by speaking in their target language.

In-game dialogue — player speaking with NPC in low-poly VR environment, Dutch speech bubbles showing conversation exchange

Voice-Based NPC Interactions

Players speak directly with NPCs using speech recognition. This core mechanic addresses the user's need for practical speaking experience, providing a safe, consequence-free environment to build confidence and experiment with contextual vocabulary.

Plagueton Summary screen — CEFR level feedback, Correctness, Delivery, and Engagement metrics with actionable feedback cards

AI-Powered Grading System

After each interaction, players receive a detailed UI breakdown of their performance based on Correctness, Delivery, and Engagement. This provides actionable, specific feedback, helping users understand exactly how to improve their conversational skills, grammar, and natural intonation.

Branching narrative flowchart — decision tree showing multiple paths, endings (Win, Loss, Misstep, Villainous), and progression states

Replayability for Proficiency

To develop real CEFR proficiency, the game uses its branching structure to turn playtime into practice. Reaching an ending encourages players to restart and explore new conversational paths. By replaying the game to master each subject and collect every item, players unlock an advanced final interaction—ensuring that their drive to complete the game directly mirrors their journey to language mastery.

The Impact

+40%

Potential vocabulary increase through NPC interactions (Rankin study)

115

Unique ways to play

94%

Task completion during usability testing

Reflection & Learnings

What I Learned

Holding my first solo workshop was instrumental; it proved that gathering diverse perspectives is crucial to stepping outside my own biases. I also learnt how to navigate Unity and the Oculus SDK, translating 2D UI concepts into tangible VR interfaces.

What I'd Do Differently

I would proactively strive to create a less Eurocentric prototype by arranging focus groups with diverse communities to challenge my biases. Additionally, better time management during the ideation phase would have allowed me more time to refine the game's branding and explore backend machine learning concepts further.