How USAID is Transforming Development with AI

USAID's AI Action Plan represents a commitment to harnessing artificial intelligence in a responsible, effective manner across a range of global development sectors. This strategy rests on three pillars: 

1. Committing to responsible, ethical, and equitable use of AI in programming;
2. Strengthening ecosystems through data infrastructure investments, capacity strengthening, and AI-ready workforce development; and
3. Forming strategic partnerships to shape a global responsible AI agenda.

The plan aims to make AI accessible and beneficial to underserved communities, enhancing program effectiveness while adhering to USAID's vision of fostering inclusive, sustainable development worldwide. Through this approach, USAID recognizes opportunities, limitations, and tensions presented by AI as a digital solution while integrating new approaches, voices, and backgrounds into developing AI systems. Most importantly, the Agency makes inclusion and equity a priority rather than an afterthought — signaling the ways responsible AI can help advance the Agency's priorities by leading the way toward a more distributed digital future. 

To combat the increasing threat of illegal logging and its contribution to deforestation in the Amazon, in 2015 Rainforest Connection (RFCx) embarked on an environmental conservation project in the State of Pará in northern Brazil, home to the Tembé tribe. Supported by USAID's Development Innovation Ventures (DIV) from 2017-2021, this initiative focused on using RFCx's innovative AI-driven audio monitoring system to protect the Tembé tribal lands. 

The technology detects the sounds of chainsaws and other logging equipment, enabling a rapid response to prevent deforestation. This project sought to protect the  Amazon trees while supporting the rights and livelihoods of the indigenous Tembé community. By leveraging cutting-edge technology with strategic partnerships and support, RFCx demonstrated an effective approach to combating environmental degradation and promoting sustainable practices. This was particularly important in 2018 and 2019  when the Amazon faced its highest deforestation rates since 2008.

This initiative highlights Action 1 of USAID's AI Action Plan, demonstrating the power of AI in enhancing environmental conservation efforts and addressing critical ecological challenges while doing so responsibly and in concert with local communities. Reflecting Action 3 of the Agency’s Action Plan, a multi-stakeholder effort between USAID, several levels of state and federal government, and civil society groups, the inclusion of a wide variety of organizations enhanced the project’s credibility and allowed partners to earn the trust of the Tembé people.

Through early support from  USAID’s Development Innovation Ventures and the Digital Development for Feed the Future initiative, Apollo Agriculture is revolutionizing farming in Kenya and Zambia. Using satellite imaging and machine learning to guide agriculture-focused lending decisions, Apollo determines which smallholder farmers are likely to repay loans while reserving a small fraction of their portfolio for high-risk lending. By analyzing farm-level data to predict credit repayment behavior, Apollo can provide smallholder farmers with loans and customized advice while also ensuring tailored and inclusive financial services in-country. This approach aids farmers in adapting to climate change, enhances food security, reduces poverty, and promotes financial inclusion in rural communities.

Apollo Agriculture aims to address the financial and informational gaps that smallholder farmers often face. Their use of AI provides farmers with the financing they need to  buy better products, increase their harvest, and turn their subsistence farming into commercial farming. The company’s work in this area  demonstrates the potential for AI to revolutionize traditional sectors.  It highlights the importance of strategic partnerships and funding in bringing such transformative technology to lower the barriers to market for smallholder farmers.

This project demonstrates a responsible approach to AI, as described in Actions 1 and 2 of USAID’s AI Action Plan, by strengthening local data ecosystems and supporting Apollo’s diverse team covering 11 countries while also responsibly adopting AI to solve pressing challenges.

In Mexico, the Secretariat of Education of the State of Guanajuato (SEG)’s Early Action System for School Permanence (SATPE), an AI-based early alert system, was developed to improve school retention and graduation rates. However, Itad, in partnership with Women in Digital Transformation (WinDT), PIT Policy Lab, and Athena Infonomics, identified a critical gender bias within SATPE which would prevent the model from accurately identifying up to 4 percent of at-risk girls who were in jeopardy of interrupting their studies. Using IBM’s AI Fairness 360 tool, the team analyzed anonymized data from the system provided by the Secretariat to detect and address these biases, ensuring the AI model effectively identifies at-risk students without gender discrimination.

This project was chosen as a winner of the USAID's Equitable AI Challenge, which aimed to address gender biases in AI systems in low- and middle-income countries. Focused on the educational sector, Itad's initiative in Mexico sought to mitigate gender biases in AI tools that could inadvertently impact school dropout rates. The project team's efforts included not only technical adjustments to the AI system but also capacity-building workshops for teachers and local government officials, emphasizing the need to consider gender and other social factors in AI development and deployment. The project propelled Itad and partners to develop an Ethical Guide and Checklist to ensure policymakers in Guanajuato understood the risks of AI. These learnings were later presented to stakeholders from the Ministry of Education in the state of Tamil Nadu, India, to explore how lessons learned from the Mexico experience could transfer to the Indian context.  The Ethical Guide and Checklist are now freely available to civil servants around the world considering integrating AI tools into their systems.

Itad's project in Mexico exemplifies Actions 1,  2, and 3 of the USAID AI Action Plan by addressing the challenge of AI fairness and bias reduction, particularly in terms of gender. The project demonstrates the importance of integrating social science and gender expertise into AI initiatives — highlighting the need for comprehensive approaches that combine technical solutions with awareness and training. The project’s post-project knowledge exchange events  brought in government stakeholders from Mexico, Argentina, Uruguay, and India, highlighting the initiative’s global approach to improving the decision-making process of partner governments looking to implement AI solutions.

Through the support of USAID, the Global Engineering and Research (GEAR) Center at MIT has developed a revolutionary low-cost precision irrigation controller, marking a significant advancement in agricultural technology. This innovaative  irrigation system uses artificial intelligence to predict local weather, enabling the system to optimize water and energy use. This caters specifically to the unique challenges faced by smallholder farmers in water-stressed regions. By predicting solar exposure and tailoring irrigation needs for individual locations, this technology improves agricultural efficiency, addressing critical issues of soil degradation and water scarcity in regions that  benefit the most from these gains.

GEAR Center's initiative responds to the urgent global need for sustainable farming practices, especially in areas where traditional methods are becoming unsustainable. With USAID's support, the lab focuses on empowering smallholder farmers around the world by providing them  access to advanced, affordable technology. Field tests of the irrigation controller in diverse locations, including in the south of Morocco, the Jordan Valley, and Puma Springs Farm in Kenya, underscore the lab's commitment to practical, on-the-ground solutions that address real-world agricultural challenges.

This work highlights a responsible approach to AI, as outlined in Action 1 of the AI Action Plan, by generating evidence for how AI tools can strengthen agricultural practices to combat drought and food insecurity.

Data Source: Food and Agriculture Organization of the United Nations. (2022). Historic Agricultural Drought Frequency Maps. Imagery Layer by giews_hqfao. Living Atlas. Updated July 17, 2023.

Historic Agricultural Drought Frequency Maps depict the frequency of severe drought in areas where 30 percent/50 percent of the cropland/grassland has been affected. The historical frequency of severe droughts (as defined by ASI) is based on the entire ASI times series (1984-2022). Formula: The number of years when land affected>30 percent/50 percent occurred/(2022-1984+1) *100

To achieve and sustain HIV epidemic control, the Innovative Data Methods for Market Segmentation of HIV Services Challenge incorporated AI to help identify groups of people living with HIV/AIDS who were willing to pay for private HIV services. Fraym and Palindrome Data, the challenge winners, developed sophisticated AI models capable of processing and analyzing vast datasets to better understand unique trends and behaviors of patients to better guide targeted HIV treatment interventions. Leveraging advanced AI methodologies, including data analytics and machine learning, the Challenge enabled more precise market segmentation. It informed decision-making among private-sector and government partners seeking better HIV treatment services. 

The Innovative Data Methods for Market Segmentation of HIV Services Challenge, launched in 2021 to support The President's Emergency Plan for AIDS Relief’s (PEPFAR) mission, aimed to tackle critical obstacles in HIV/AIDS treatment services. The Challenge was launched to identify people living with HIV/AIDS who are willing and able to pay for private HIV services. This practice, known as market segmentation, is vital for tailoring HIV treatment services and expanding market diversification— essential steps towards preventing future HIV infections, reducing the number of deaths as a result of HIV/AIDS, and providing better care to people living with HIV. USAID’s Bureau of Global Health played a pivotal role in this initiative by funding the Challenge winners, Fraym and Palindrome Data, with a grant to conduct market segmentation analyses in priority PEPFAR countries like Kenya and South Africa.

This initiative highlights Action 1 of the AI Action Plan by leveraging innovative data analytics and AI methodologies to address a significant health issue while embracing the responsible use of AI for development programming.

In partnership with USAID, the Wadhwani Institute for Artificial Intelligence is deploying AI solutions to tackle the escalating challenge of tuberculosis (TB) in India, particularly the rise in drug-resistant strains. The Transformative Research and Artificial Intelligence Capacity for Elimination of Tuberculosis and Responding to Infectious Diseases  (TRACE-TB) project exemplifies this effort, utilizing AI to enhance TB detection, treatment, and management. This initiative comes at a critical time, coinciding with India’s National TB Elimination Programme's shift to case-based reporting, a system where individual tuberculosis cases are reported and tracked on a case-by-case basis — producing an extensive health data set ideal for AI analysis. Wadhwani AI’s strategy involves the development of AI tools that analyze this data to improve early detection and treatment adherence (loss to follow-up), significantly aiding healthcare professionals in TB management.

Image: An example of an AI-based risk stratification mobile application for field staff, enlisting TB patients with higher risk for adverse outcomes who can be provided intensive support and interventions.

Wadhwani AI is recognized for its focus on developing AI solutions for social good and is collaborating with USAID on the TRACE-TB project. This partnership focuses on tackling  TB in India, where the disease poses a significant challenge.  The project's emphasis on AI stems from the need for advanced interventions to manage the evolving dynamics of TB, including drug-resistant forms.

The TRACE-TB project aligns with Action 1 of the AI Action Plan by using AI to enhance healthcare outcomes and address infectious diseases effectively.  The focus on ethical AI use, ensuring data privacy and cultural sensitivity, is a testament to the project's commitment to responsible AI practices.

Data Source: Ministry of Health and Family Welfare, Government of India. (2022). INDIA TB REPORT 2022. 3.3 Treatment outcome of TB patients notified in 2020 (Total).

The Machine Learning for Peace Project (MLP), in collaboration with USAID’s Bureau for Democracy, Human Rights, and Governance (DRG) and various partners, leverages machine learning and natural language processing to address the critical global challenge of shrinking civic spaces. Emerging from the need identified by the Legal Enabling Environment Program, MLP focuses on the real-time, proactive monitoring of civic space trends. The project's advanced data collection and analysis capabilities enable the provision of early warnings for potential closures of civic spaces, empowering stakeholders with the tools to anticipate and react effectively.

The MLP initiative, part of the Illuminating New Solutions and Programmatic Innovations for Resilient Spaces (INSPIRES) consortium and developed with the DRG Bureau and various partners, is designed to address the growing challenge of closures in civic space globally. Utilizing advanced techniques in natural language processing and machine learning, MLP focuses on collecting big data to forecast changes in civic spaces. This approach provides early warnings, enabling stakeholders to better prepare for and respond to these changes. By predicting and preventing restrictions on civil society, and offering support through technical assistance and subgrants, MLP represents a significant step forward in protecting democratic freedoms and bolstering civil society, especially in challenging environments.

MLP's application of AI in monitoring and predicting changes in civic spaces aligns with USAID's AI Action Plan, particularly in its emphasis on employing advanced technology for promoting democracy and human rights. The initiative's approach to leveraging AI for forecasting and prevention demonstrates a commitment to using technology responsibly to support civil society.