AarogyaSathi: Design and Preliminary Evaluation of a Multilingual Public Health Information Chatbot

AarogyaSathi is a deployed web prototype with a demon-strated WhatsApp test-number interface that combines curated health information, multilingual retrieval, claim checking, emergency rules, and experimental geographic signals. We reproduced the supplied component benchmarks successfully, but training overlap and configuration differences limit how far they can be interpreted. The live synthetic conversations exposed an emergency-negation error, unrelated emergency advice, and a failed Odia answer, and source inspection found faulty vaccination age mapping. These findings give us concrete priorities for improving the system. The What-sApp follow-up established basic bidirectional delivery for an approved test recipient.

[1]     World Health Organization et al., “Managing the COVID-19 infodemic: Promoting healthy behaviours and mitigating the harm from misinformation and disinformation,” Sep. 23, 2020. [Online]. Available: https://www.who.int/news/item/23-09-2020-managing-the-c ovid-19-infodemic-promoting-healthy-behaviours-and-mitigating-the-h arm-from-misinformation-and-disinformation. Accessed: Oct. 29, 2026. [2]     DRIEMS University, “Smart India Hackathon problem statements,” SIH25049, Government of Odisha, 2025. [Online]. Available: https://www.driems.ac.in/wp-content/uploads/2025/09/Smart-India-Hac kathon_PROBLEM-STATEMENTS.pdf. Accessed: Oct. 29, 2026. [3]     H. L. Semigran, J. A. Linder, C. Gidengil, and A. Mehrotra, “Evaluation of symptom checkers for self diagnosis and triage: Audit study,” BMJ, vol. 351, Art. no. h3480, 2015, doi: 10.1136/bmj.h3480. [4]     P. Lewis et al., “Retrieval-augmented generation for knowledge-intensive NLP tasks,” in Advances in Neural Information Processing Systems, vol. 33, 2020. [Online]. Available: https://proceedings.neurips.cc/paper/2020/hash/6b493230205f780e1bc 26945df7481e5-Abstract.html [5]     F. Feng, Y. Yang, D. Cer, N. Arivazhagan, and W. Wang,“Language-agnostic BERT sentence embedding,” in Proc. ACL, 2022,pp. 878–891, doi: 10.18653/v1/2022.acl-long.62. [6]     L. Laranjo et al., “Conversational agents in healthcare: A systematic review,” Journal of the American Medical Informatics Association, vol. 25, no. 9, pp. 1248–1258, 2018, doi: 10.1093/jamia/ocy072. [7]     M. Milne-Ives et al., “The effectiveness of artificial intelligence conversational agents in health care: Systematic review,” Journal of Medical Internet Research, vol. 22, no. 10, Art. no. e20346, 2020, doi: 10.2196/20346. [8]     National Health Mission, Ministry of Health and Family Welfare, Government of India, “National Immunization Schedule.” Accessed: Oct. 29, 2026. [Online]. Available: https://nhm.gov.in/New_Updates_2018/NHM_Components/Immunizat ion/report/National_%20Immunization_Schedule.pdf [9]     National Centre for Disease Control, Government of India, “Integrated Disease Surveillance Programme.” Accessed: Oct. 26, 2026. [Online]. Available: https: //ncdc.mohfw.gov.in/includes/About/CentresAndDivision/IDSP.php [10]    World Health Organization, Ethics and Governance of Artificial Intelligence for Health: Guidance on Large Multi-Modal Models, 2024, ISBN 978-92-4-008475-9. [Online]. Available: https://www.who.int/publications/i/item/9789240084759. Accessed: Oct. 29, 2026.