Introduction
The Agency for International Development leverages artificial intelligence across a diverse range of initiatives to address global challenges and improve developmental outcomes. From monitoring social media for misinformation to optimizing water resource management using satellite data, AI applications are enhancing the efficiency and effectiveness of development efforts. Additionally, AI is used to predict treatment interruptions and improve healthcare resource management. The following AI use cases are transforming data analysis, resource management, and community support, underscoring the Agency’s commitment to innovative problem-solving and sustainable development.
Use Cases
-
1. Media Early Warning System (MEWS)
The Media Early Warning System (MEWS) is designed to monitor social media for alterations in images and videos that may indicate malign narratives. By detecting emerging trends and narratives, MEWS aims to counteract misinformation and harmful content, thereby enhancing the ability to respond to potential threats to public perception and social stability.
-
2. Long-term impacts of land-use/land-cover dynamics on surface water quality in Botswana’s reservoirs using satellite data and artificial intelligence methods: Case study of the Botswana’s Limpopo River Basin (1984-2019)
This comprehensive study examines the long-term impacts of land-use and land-cover (LULC) dynamics on surface water quality in Botswana’s Limpopo River Basin (LRB) from 1984 to 2019. Given that semi-arid Botswana relies heavily on its reservoirs, which are vulnerable to LULC changes and runoff, the research aims to understand the intricate relationships between land and water resources. By employing data-driven artificial intelligence methods, the study will quantitatively analyze how LULC changes, socioeconomic development, and climate change affect water quality and availability. The research will also predict future scenarios for 2020-2050. To enhance data acquisition for LULC analysis, the study will utilize optical Earth-observation and meteorological satellite data. Additionally, it aims to develop empirical models for continuous monitoring of reservoir water quality, using 35 years of in-situ measurements and drone-borne spectrometer observations, thereby providing a cost-effective approach to water quality management.
-
3. Morogoro youth empowerment through establishment of social innovation (YEESI) lab for problem-centered training in machine vision
The Morogoro Youth Empowerment through Establishment of Social Innovation (YEESI) Lab project aims to create a machine vision program for youth in the Morogoro region of Tanzania. Despite having knowledge in information technologies, many young people lack the skills to address machine vision challenges, particularly in agriculture. This initiative seeks to raise awareness and equip youth with the necessary skills to tackle pressing agricultural issues faced by over 80% of Tanzania’s population engaged in farming. The project focuses on five key areas: (1) Disease Detection and Classification: Training experts to identify crop and livestock diseases using machine vision. (2) Weed Classification: Developing algorithms for accurate weed identification to enhance automatic detection efforts. (3) Pest Detection and Classification: Creating machine vision tools for Integrated Pest Management (IPM) to assist farmers in pest identification and mitigation. (4) Crop Seedlings Stand Count and Yield Estimation: Utilizing machine vision and drones for accurate stand counts and yield estimation, improving replanting strategies. (5) Crop Vigor Estimation: Developing algorithms to assess crop health, enabling farmers to apply inputs more effectively and improve overall crop performance. This project aims to foster scientific societies and enhance agricultural productivity in the region.
-
4. Project Vikela
Project Vikela focuses on utilizing artificial intelligence to enhance the detection of illegal rhino horn trafficking through the analysis of luggage X-ray scans at airports. By implementing AI technologies, the project aims to improve the effectiveness of customs and border protection efforts in combating wildlife trafficking, thereby contributing to the conservation of endangered species.
-
5. Using ML for predicting treatment interruption among PLHIV in Nigeria
This project focuses on using machine learning to predict treatment interruptions among people living with HIV (PLHIV) in Nigeria. By analyzing data from the USAID-funded Strengthening Integrated Delivery of HIV/AIDS Services (SIDHAS) project, an algorithm was developed to estimate the likelihood of newly initiated patients on antiretroviral therapy (ART) interrupting their treatment. The algorithm has been integrated into the Lafiya Management Information System (LAMIS), which tracks individual patient records. Weekly outputs are shared with healthcare staff, allowing them to identify high-risk patients and provide targeted follow-up support to minimize treatment interruptions. Additionally, qualitative assessments were conducted with healthcare workers to gather insights on their perceptions of machine learning and identify further support needed for its integration into routine practices.
-
6. Breakthrough RESEARCH’s Social Media Listening
Breakthrough RESEARCH’s Social Media Listening initiative employs machine learning to analyze and organize large volumes of data shared on social media platforms. The project focused on 12,301 social media posts in Nigeria to investigate the evolution of gender-related conversations over the past five years. Utilizing Crimson Hexagon’s machine learning algorithm, “Brightview,” the project scraped publicly available content relevant to reproductive health/family planning (RH/FP) and youth. The posts were classified by topic based on detected language, resulting in a comprehensive dataset that categorizes conversations into overarching themes. This analysis enables researchers to identify key trends in conversation volume, misinformation, attitudes, and social norms, providing valuable insights that surpass traditional public health research methods.
-
7. Serbia: AI predictions for the utilization of hospital beds
In Serbia, AI technology was employed to develop a proof-of-concept model for predicting hospital bed occupancy using Ministry of Health (MoH) data from 2019. The model achieved an overall median error of approximately 20% by department, demonstrating the potential of AI in healthcare resource management. Following this initial success, the Center for Health Information Systems and Utilization (CHISU) was tasked with exploring a higher-priority use case: optimizing waiting lists for scheduled imaging diagnostics services, specifically for CT and MRI scans. This initiative aligns with the national AI strategy and aims to enhance data-driven decision-making in the healthcare sector, with plans for implementation in 2023-2024.
-
8. Mali: AI predictions for the optimization of the allocation of the distribution of COVID-19 vaccines
In Mali, AI technology was utilized to create a proof-of-concept model aimed at optimizing the allocation of COVID-19 vaccines. This pandemic preparedness model employs a multi-tiered strategy that prioritizes target populations, specifically focusing on areas with high COVID-19 case rates (hotspots) and pregnant or breastfeeding women. By leveraging data from the District Health Information Software 2 (DHIS2), the model seeks to enhance the efficiency and effectiveness of vaccine distribution during the pandemic.
-
9. Indonesia: AI predictions for improving forecasts for TB drugs
In Indonesia, AI technology is being harnessed to develop a forecasting model specifically for tuberculosis (TB) sensitive drugs. This model aims to improve the accuracy of annual quantification exercises conducted by the Ministry of Health (MoH) and is linked to the national data integration platform, SatuSehat. By providing more precise forecasts, the model will support better planning and allocation of TB medications, ultimately enhancing public health outcomes.
-
10. NASA SERVIR - Bias Correcting Historical GEOGloWS ECMWF Streamflow Service (GESS) data using Machine Learning (ML) Techniques
The NASA SERVIR initiative focuses on enhancing the GEOGloWS ECMWF Streamflow Service (GESS) by applying machine learning techniques to bias-correct historical discharge data. GEOGloWS serves as a collaborative platform for the international hydrologic sciences community, facilitating the use of Earth Observations (EO) for river flow forecasting and providing a 40-year simulated historical flow dataset. By utilizing a Long Short-Term Memory (LSTM) model, this application aims to improve the accuracy of globally available discharge information, thereby supporting better water resource management and decision-making.
-
11. NASA SERVIR - Using artificial intelligence to forecast harmful algae blooms in Lake Atitlán, Guatemala
The NASA SERVIR project in Guatemala employs machine learning techniques combined with Earth observations and weather-modeled data to forecast daily algal blooms in Lake Atitlán. This forecasting system is utilized by local authorities, including the Authority for Sustainable Management of the Lake Atitlán Basin and its surroundings (AMSCLAE), to enhance their Harmful Algal Blooms Alert System. Supported by National Geographic and Microsoft through their AI for Innovation grants, this initiative aims to improve water quality management and protect the lake’s ecosystem from harmful algal blooms.
-
12. NASA SERVIR - Mapping urban vulnerability using AI techniques
The NASA SERVIR initiative focuses on enhancing urban vulnerability assessments in critical population centers by developing replicable methods for utilizing satellite imagery to map informal settlements. This project aims to provide valuable insights into urban planning and disaster risk management, enabling authorities to better understand and address vulnerabilities in rapidly growing urban areas.