Motivation

In mobile AR, users are often overwhelmed by too much or too little information. Depending on the situation, they may need either a quick visual summary or in-depth content—but current systems rely on manually predefined rules and static designs.

This thesis explores how AI can dynamically generate context-sensitive Levels of Detail (LoD) for AR interfaces—both within a single modality (e.g., text-to-text) and across modalities (e.g., text-to-icon).

You will:

• Analyze information needs in mobile AR use cases (e.g., navigation, urban awareness, mobility assistance).
• Design methods to automatically generate LoD representations using NLP or multi-modal AI models.
• Prototype real-time LoD switching between detailed text, summaries, and visual indicators.
• Evaluate the impact on usability and cognitive load in dynamic scenarios.