A novel framework called VirtualXAI that transforms XAI evaluation by leveraging GPT-generated personas to assess how well AI explanations work for diverse users.

Key Innovations:

• Creates diverse synthetic user personas to test XAI methods across different user needs and backgrounds
• Enables systematic assessment of explanation quality from multiple user perspectives
• Provides a user-centric approach to XAI evaluation rather than purely technical metrics
• Bridges the gap between technical XAI capabilities and real-world user requirements

Security Relevance: This approach directly addresses AI trustworthiness and transparency concerns - critical for secure AI systems in sensitive domains. By ensuring explanations work for diverse users, organizations can better manage risks associated with opaque AI decision-making.

VirtualXAI: A User-Centric Framework for Explainability Assessment Leveraging GPT-Generated Personas