Abstract

Generative AI has been lately one of the turning points of Artificial Intelligence life cycle hype that can generate new content such as high-quality text, images, audio, simulations and others. Generative AI Models create an opportunity in humanitarian space for fast and efficient information retrieval in emergency responses and monitoring and predicting trends in humanitarian crises.

Evidence-based decision-making is key for humanitarian operations with the increasing rate of crises and limited resources. While there is an increased availability of expansive qualitative and quantitative data, the vast amount of unstructured and convoluted data in complex settings creates an urgent need for responders to apply fast and real-time actionable data. The analysis needs to be presented through easy-to-use and collaborative tools that consider the emergent nature of crises and the capacity required to analyse large datasets.

To address this, DFS has been building an accessible, AI-powered “virtual data assistant” for humanitarian staff. This tool will enable them to access, extract, and visualise life-saving information from the vast data sources available publicly and within responding organisations. Our main objectives are developing an up-to-date structured humanitarian response knowledge base to elevate data access and collaboration; provide augmented access to near real-time data insights to empower both technical and non-technical humanitarian responders to effortlessly recognize patterns and correlations; and elevate the capability for rapid emergency response and anticipation among local and international organisations.

GANNET (Guided Assistance and Navigation for Needs and Emergency Tasks) web application will combine the power of a virtual AI assistant, an up-to-datehumanitarian knowledge hub, deep insights, and robust security to revolutionise how front-line humanitarian organisations access and react to sector-specific information. The innovation and novelty of GANNET is in the development of a continuously updated search-query system through a Generative AI assistant that understands natural language queries and provides near real-time, bias-aware, and high-quality responses in the context of the humanitarian/development space. The platform will equip humanitarian responders with easily identifiable event trends and relationships in historical and live data to help predict developments in the field.

In this technical session, we will explain the architecture behind GANNET, explore use cases together with professionals from international and local organisations, and how this generative-AI-based platform can support operational and strategic decision-making. We will introduce the demo of GANNET and receive feedback from the audience to continuously respond to humanitarian needs.

The targeted audience will be organisations who are interested in exploring a practical implementation of generative AI in the humanitarian space and how to collaborate on the use of such AI tools both in local and international complex settings.