LiteVR: Interpretable and Lightweight Cybersickness Detection using Explainable AI

Eye-tracking is the key indicator of cybersickness: The research found that eye-tracking features are the most dominant predictors for detecting cybersickness in VR, suggesting that monitoring how users' eyes move and respond can reliably signal when someone is experiencing discomfort in virtual environments.

AI transparency improves trustworthiness for VR deployment: By using explainable AI techniques (SHAP), the researchers made their cybersickness detection models interpretable rather than "black-box," addressing a critical barrier to deploying AI systems on consumer VR devices where users need to understand why the system is making decisions about their experience.

Responsible AI doesn't require sacrificing performance: The framework achieved significant efficiency gains—reducing model size by 4.3x, training time by 5.6x, and inference time by 3.8x—while actually improving detection accuracy, demonstrating that making AI systems more lightweight and explainable doesn't mean compromising on results.

Practical AI requires considering real-world constraints: The research addresses a often-overlooked responsibility in AI development: ensuring systems can actually run on the devices they're meant for, rather than requiring expensive servers, making cybersickness detection accessible on standalone VR headsets with limited computing power and battery life.