HIBA Project Post HEI – October 2024
The American University of Beirut's Humanitarian Engineering Initiative (HEI) has completed an innovative project titled “Health Information Bot Assistant – HIBA: AI-mediated Interactive Health Messaging for Community Health Promotion in Low-Middle Income Countries." Funded by the Bill & Melinda Gates Foundation - Grand Challenges Grants, this project leverages artificial intelligence to bridge the gap in health communication, providing reliable health information to underserved communities with varying literacy levels.
By harnessing the potential of Artificial Intelligence (AI), this enhanced platform allows for the timely and adequate provision of health information to vulnerable groups, who mostly encounter systemic and structural obstacles in obtaining health care information. The platform is able to provide fully interactive, real-time, reliable and accurate responses to the community health questions and concerns.
The technology was developed based on a conducted context assessment in Lebanon and Jordan, in collaboration with the International Federation of Red Cross and Red Crescent Societies (IFRC), the Lebanese Red Cross (LRC) and the Jordan Red Crescent (JRC), to identify the target audiences' needs, available resources and health literacy factors of importance. Accordingly, through ReAct prompt engineering, GPT4 integration and data processing, the chatbot was trained to control accuracy of responses and minimize the risk of hallucinations. Additionally, multilingual speech-to-text and text-to-speech technologies, along with a geographically-determined multilingual conversational flow, were integrated to further increase the chatbot's accessibility to varying literacy levels in the target communities. Cloud resource management, performance monitoring, and infrastructure provisioning were also maintained to sustain the chatbot's optimal operation. Subsequently, a 4-level testing framework was created to evaluate the chatbot's performance and responses, and to incorporate feedback into the continuous development and optimization. A combination of corpus validation, dashboard metrics analysis, and multi-phase user satisfaction qualitative assessment and thematic analysis were conducted.
The chatbot's key features, specifically designed to mitigate literacy-mediated barriers, include:
1. Multilingual Support: By supporting 28 languages, the chatbot can interact with users in their native languages, making health information more accessible and comprehensible.
2. Speech-to-Text and Text-to-Speech Technology: To cater to varying literacy levels, the chatbot includes human-like speech-to-text and text-to-speech functionalities. This feature ensures that even individuals with limited reading and writing skills can access and benefit from the health information provided.
3. Geographically-Determined Multilingual Conversational Flow: The chatbot's responses are optimized based on geographical locations, ensuring relevance and appropriateness in the local context.
4. ReAct and Prompt Engineering: Advanced AI techniques, including ReAct and prompt engineering, have been employed to enhance the chatbot's interactive capabilities, ensuring coherent and contextually appropriate conversations.
5. Ensured Quality and Reliability: Extensive quality assessment, validation and optimization of the health responses were conducted, which involved building a robust multi-lingual database through corpus generation of prevalent health topics of interest to the communities (determined through FGDs), and content scraping from national and regional reliable resources.
6. Dashboard for Usage Analytics and Performance Tracking: A dashboard was also developed for metrics evaluation, allowing the team to monitor the chatbot's performance and user engagement. This data-driven approach ensured continuous improvement and adaptation to meet the evolving needs of the target communities.
As of October 2024, amid the ongoing war in Lebanon and resulting displacement of millions of people, HIBA stands as a transformative tool in the realm of public health communication. By deploying the chatbot in displacement centers throughout Lebanon using designed posters with a QR Code, HIBA's services would be accessible to all IDPs. Ultimately, reliable health information would be provided to underserved communities with varying literacy levels.
The chatbot's ability to provide real-time, reliable health information in multiple languages, coupled with its speech-to-text and text-to-speech functionalities, makes it a powerful tool for health promotion across all demographics and literacy levels.