Explainable AI: Feature Attribution & Privacy Simulator

Machine-learning models used for high-stakes decisions — credit approval, hiring, medical triage — often behave as opaque "black boxes." This simulator visualizes the explainable-AI techniques used to open that box: a central model core takes four applicant features and produces an approve/deny decision, while a SHAP-style attribution layer shows exactly how much each feature pushed the outcome one way or the other. Adjust the feature sliders to see attributions recompute in real time, then dial the differential-privacy budget ε down to see how privacy-preserving noise degrades the reliability of the very explanation you're being shown — the central tension between interpretability, reliability and safety in modern ML systems.