PriXAI
Explainable AI for privacy-aware decision support
Duration: April 2026 - March 2031
Funding: NWO Vidi Grant
Imagine trusting an AI system to support decisions in healthcare, where mistakes can cost lives. Today, many AI explanations remain inconsistent, offer little guidance for use, and can even leak private information.
PriXAI tackles this problem by shifting the focus away from reverse-engineering model internals and toward a more practical question: when and why should we trust an AI prediction? The project develops methods that provide clear, actionable evidence while preserving privacy.
By combining data-centric AI with privacy-preserving learning, PriXAI aims to build AI systems that people can trust in practice. It reimagines explainability not just as a source of insight, but as a foundation for action and trust.
Team
- Megha Khosla (Lead)
- Hanwen Ge (PhD student, starting June 1, 2026)
- Giray Düzel (PhD student, starting September 1, 2026)