Skip to main content Skip to navigation

SUKHSANDESH: An Avatar Therapeutic Question Answering Platform for Sexual Education in Rural India

Project Overview

The document discusses S UKH SANDESH, an innovative AI-driven Question Answering platform designed to deliver sexual education to vulnerable rural communities in India. Utilizing generative AI and avatar technology, it effectively tackles cultural stigmas surrounding sexual health by providing accessible information in regional languages. The platform prioritizes user safety through protective measures that ensure privacy and mitigate exposure to harmful content. Furthermore, the empathetic avatar feature enhances user engagement, particularly benefiting individuals with limited literacy skills by fostering a more relatable and supportive learning environment. Overall, the application of generative AI in this educational context demonstrates its potential to improve access to vital information and promote informed decision-making among underserved populations.

Key Applications

S UKH SANDESH - AI-based Question Answering platform

Context: Rural Indian population seeking sexual education

Implementation: Multi-staged system using information retrieval and generative AI (LLMs) with avatar therapy for real-time audio responses.

Outcomes: Improved access to sexual education, enhanced user engagement through empathy, and support for regional languages.

Challenges: Cultural stigma around sexual education, lack of labeled data, ensuring data quality, and ethical considerations regarding user privacy.

Implementation Barriers

Cultural Barrier

Stigmas and taboos surrounding sexual education in rural India hinder open communication and data acquisition due to cultural norms related to sexual topics.

Proposed Solutions: Ground-level awareness programs and partnerships with local organizations (e.g., Gram Vaani) to build trust, along with collaboration with local organizations to gather anonymized user queries and responses.

Technical Barrier

Quality issues in data due to potential biases, lack of representation in training datasets, and a shortage of skilled personnel in AI for model training and deployment.

Proposed Solutions: Implement rigorous data validation processes, utilize diverse data sources, and invest in training programs to build local AI expertise.

Resource Barrier

High energy consumption associated with training large language models leading to a significant carbon footprint.

Proposed Solutions: Explore energy-efficient AI models and practices to mitigate environmental impact.

Project Team

Salam Michael Singh

Researcher

Shubhmoy Kumar Garg

Researcher

Amitesh Misra

Researcher

Aaditeshwar Seth

Researcher

Tanmoy Chakraborty

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Salam Michael Singh, Shubhmoy Kumar Garg, Amitesh Misra, Aaditeshwar Seth, Tanmoy Chakraborty

Source Publication: View Original PaperLink opens in a new window

Project Contact: Dr. Jianhua Yang

LLM Model Version: gpt-4o-mini-2024-07-18

Analysis Provider: Openai

Let us know you agree to cookies