Digital Upstander XR

AI-XR-based adaptive ethics training game. Going beyond the limits of static cyber ethics education, fostering 'active Upstanders' through the real-time adaptivity of a Large Language Model (LLM).

Comment

Inspiration

We sought to move digital citizenship education beyond lectures. Inspired by the strict EU DSA/BIK+ guidelines, our goal was to create an immersive 3D simulator where students face real-time deepfake and denigration scenarios, forcing immediate ethical decision-making.

What it does

Digital Upstander XR is an AI-driven ethical training game. It dynamically presents escalating scenarios, captures the player's choice (A, B, C, D), and instantly calculates consequences (Badges, Digital Wisdom, Empathy) to deliver immediate, gamified, and pedagogically sound feedback.

How we built it

We engineered a robust Unity-to-LLM pipeline: Unity (C#): The client handles the 3D environment and UI. Custom scripts (N8nWebhookClient.cs) send player choices as JSON Requests. n8n Middleware: Acts as the backend, receiving the request, applying Scoring Rules, and orchestrating the AI. LLM (AI Agent): Generates all dynamic dialogue, situation summaries, and scores, ensuring alignment with the required EU pedagogical styles.

Challenges we ran into

Unity Complexity: Overcoming initial struggles with Unity UI component referencing and connecting complex script instances on the same GameManager object. Gamification Balance: Integrating the strict educational rules (Good/Bad/Moderator Badges) into a motivating feedback loop without being overly punitive. Network Synchronization: Managing asynchronous communication delays (using C# Coroutines) to ensure a smooth, non-blocking flow between Unity and the external n8n server.

Accomplishments that we're proud of

Successful LLM Integration: Built a robust pipeline allowing the AI to drive the narrative in a real-time 3D environment. Modular Design: Created reusable, delegated scripts for easy scalability to new scenarios (e.g., Cyberbullying). Educational Alignment: Scoring is directly based on the EU BIK+ and DSA guidelines (Identify, Protect, Report, Rebuild).

What we learned

We confirmed the critical role of n8n middleware for complex LLM logic and mastered the intricate process of Unity Inspector component linking—a key to robust game architecture

What's next for Digital Upstander XR

Enhance Immersion: Implement LLM-triggered Character Animation and Dynamic Atmosphere changes based on emotion codes. Data Sustainability: Integrate Firebase Firestore for reliable, multi-session Score Persistence. Business Growth: Launch a SaaS subscription model for educational institutions, emphasizing EU compliance and offering tiered premium content modules.

Built With

Share this project:

Updates