MUMBAI, India, June 30 -- Intellectual Property India has published a patent application (202641053820 A) filed by Bastin K J on April 28, 2026, for "thozhi:a Safety Layered Trigger Activated Chatbot For Mental Health Pre-Diagnostic Screening".
Inventors include Bastin K J; Anita Jasmine R; Aadithya Anand; and Tarun Dhakshinamoorthy.
The application for the patent was published on June 26, 2026, under issue no. 26/2026.
Abstract: ABSTRACT The rapid integration of Large Language Models (LLMs) into the mental health landscape has created a doubleedged sword: while accessibility is at an all-time high, so are the risks of "hallucinated" advice or clinically unsafe interactions. This paper introduces Thozhi, a hybrid Al framework designed to move beyond unconstrained chat and toward a safer, ethically grounded model for mental health screening. At its core, the architecture employs a persona-controlled LLM paired with a semantic distress detection layer. Unlike traditional bots that either aimlessly converse or immediately bombard users with forms, Thozhi remains in a supportive conversational state until specific distress indicators are met. Only then does the system trigger validated medical instruments, specifically the Patient Health Questionnaire-9 (PHQ-9) and the General Anxiety Disorder-7 (GAD-7). This "conditional activation" strategy ensures that clinical rigor is applied only when necessary, effectively reducing user fatigue and the mechanical feel of standard screenings. Our evaluation shows that Thozhi achieves higher trigger precision and significantly improves user-perceived empathy compared to baseline models. By reframing the LLM as a structured decision support bridge rather than an autonomous diagnostic authority Thozhi provides a responsible, clinically informed roadmap for the future of digital mental health support.
Disclaimer: Curated by HT Syndication.
