Intellectual Property India Publishes Patent Application for Hybrid Explainable Ai System For Real-Time Decision Transparency With Multi-Model Fusion And Adaptive Explanation Generation

MUMBAI, India, June 30 -- Intellectual Property India has published a patent application (202641075291 A) filed by Dr. Sujatha S R; Dr. Raviram V; Sindhu Tumkur Nagaraj; Madhumala Gugutaiah; and Dr. Rashmi H C on June 18, 2026, for Hybrid Explainable Ai System For Real-Time Decision Transparency With Multi-Model Fusion And Adaptive Explanation Generation.

Inventors include Dr. Sujatha S R; Dr. Raviram V; Sindhu Tumkur Nagaraj; Madhumala Gugutaiah; and Dr. Rashmi H C.

The application for the patent was published on June 26, 2026, under issue no. 26/2026.

Abstract: ABSTRACT OF THE INVENTION Title: Hybrid Explainable AI System for Real-Time Decision Transparency with Multi-Model Fusion and Adaptive Explanation Generation The invention relates to a hybrid artificial intelligence system that provides real-time decision transparency by combining high-accuracy black-box models with interpretable models through intelligent fusion mechanisms. The system comprises an input layer (100) with data ingestion (110), feature engineering (120), and context encoder (130); a hybrid AI engine (200) with black-box models (210-213), interpretable models (220-223), hybrid fusion (230), confidence estimator (240), and decision output (250); an explainability engine (300) with local explanations (310), global explanations (320), attention maps (330), and reasoning trace (340); a real- time transparency layer (400) with explanation generator (410), NLG module (420), visual explanation (430), latency optimizer (440), and stakeholder adapter (450); an output interface (500) with dashboard (510), API (520), audit logger (530), and report generator (540); and a feedback module (600) with human feedback collector (610), model monitor (620), and continuous learning (630). The present invention ensures simultaneous high-accuracy predictions and human-understandable explanations through confidence-based routing between model types, parallel multi-modal explanation generation (LIME, SHAP, attention visualization, reasoning traces), stakeholder-adapted presentation, and continuous improvement via human feedback, suitable for high-stakes applications in healthcare, finance, legal, and autonomous systems requiring regulatory compliance and human oversight.

Disclaimer: Curated by HT Syndication.