A Trustworthy By Design Classification Model for Building Energy Retrofit Decision Support

AI can help solve the building energy crisis while meeting regulatory requirements: The framework successfully demonstrates that machine learning can automate retrofit recommendations for residential buildings—a critical climate priority—while complying with EU AI regulations and ethics guidelines, addressing a key gap in responsible AI deployment for environmental applications.

Transparency tools are essential for building trust in AI recommendations: The researchers used SHAP (SHapley Additive exPlanations) to explain why the system recommends specific energy efficiency strategies, making the AI's decision-making process understandable to users rather than treating it as a "black box"—a crucial requirement for responsible AI in high-stakes decisions.

Limited data doesn't have to limit AI solutions: By using a technique called CTGAN to artificially generate additional training data, the system could work effectively even with small or imbalanced datasets (demonstrated in the Latvia case study), showing that responsible AI development is possible even when real-world data is scarce.

"Trustworthy by design" requires planning from the start: Rather than adding safeguards after building the system, the researchers built transparency, explainability, and regulatory compliance into the framework from the beginning, demonstrating that responsible AI practices should be foundational rather than afterthoughts.

Real-world validation across different contexts matters: The framework was tested on both a large English dataset and a smaller Latvian dataset, showing that responsible AI systems must prove they work reliably across diverse real-world scenarios before deployment.