MUMBAI, India, Sept. 28 -- Intellectual Property India has published a patent application (202621100972 A) filed by Mit Adt University on August 20, 2026, for A Dynamic Trust State Monitoring And Response Adaptation Framework For Conversational Agents.
Inventors include Kanchan Jadhav Katake; and Dr. Jagannath Nalavade.
The application for the patent was published on September 25, 2026, under issue no. 39/2026.
Abstract: The present invention relates to an AI-based Dynamic Trust State Monitoring and Response Adaptation Framework (100) for conversational agents. The framework comprises a Text Encoding and Preprocessing Layer (101), Multi-Modal Fusion Engine (102), Trust-Aware Self-Attention Module (103), Conversation Memory Module (104), Hierarchical Attention Module (105), Multi-Task Learning Module (106), Adaptive Loss Weighting Module (107), Trust Prediction Module (108), Emotion Prediction Module (109), Engagement Prediction Module (110) and Response Adaptation Module (111). The framework processes linguistic, temporal and behavioural information associated with multi-turn conversations. A dynamically learned trust relevance score is applied to conversational elements through trust-aware self-attention, while an LSTM-based conversation memory models longitudinal trust evolution. Multi-modal fusion combines textual, temporal and behavioural representations, and hierarchical attention captures turn-level and conversation-level trust dependencies. Trust, emotion and engagement are jointly predicted using multi-task learning with adaptive loss weighting. The resulting dynamic trust state is supplied to a response adaptation module for modifying conversational responses according to detected trust, emotion and engagement conditions. The framework thereby provides continuous, multi-modal and interpretable monitoring of conversational trust and adaptive response generation.
Disclaimer: Curated by HT Syndication.
