Presentations

Abstract

As AI-driven monitoring technologies become more prevalent, ensuring their reliability and transparency is critical. Too often, AI models are perceived as obscure, raising concerns about their precision and decision-making processes. However, AI is only as effective as the methods used to validate it. This talk presents a structured approach to performance validation that integrates expert opinion with data science methodologies, demonstrating that AI can be transparent and trustworthy when evaluated correctly. Our approach combines two complementary validation techniques: ground truth with domain experts and quantitative and qualitative performance evaluation using data science methods.