How could AI be used to improve development?

If you ask online chatbot ChatGPT how artificial intelligence could be used in global development, it lists a variety of ways: predictive analytics for disease outbreaks and natural disasters; computer vision for urban planning; automated monitoring and evaluation for impact assessment; and smart resource management for agriculture. Experts agree it can do all this and more, with the potential to accelerate progress toward the Sustainable Development Goals. “AI has the potential to transform a lot of the aspects of humanitarian and development work as we know it,” said Bernhard Kowatsch, head of the innovation accelerator at the World Food Programme. On an organizational level, it can streamline operations and optimize data insights while also making development programs more accessible to communities, he said. AI, according to IBM, is the leveraging of computers and machines “to mimic the problem-solving and decision-making capabilities of the human mind.” The first theory of AI emerged in the 1950s when British mathematician Alan Turing began exploring whether machines could think. Since then, it’s morphed into many things: facial recognition tools, map applications, and, more recently, language models such as ChatGPT. So far, such innovations have been deployed mainly by the private sector, said Vilas Dhar, president of the Patrick J. McGovern Foundation, which explores how advances in information and technology can improve quality of life. “But we'd be failing our social mission if we didn't bring these tools to bear on the greatest challenges facing humanity,” he added. “The best version of an AI-driven future is one that's informed and shaped by the humanitarian sector's experience [in] deploying, evaluating, and iterating on these tools.” With that in mind, Devex asked experts how AI is already being used in the global development and humanitarian sectors. Disaster response When a crisis hits, a rapid response is imperative to prevent further damage and loss of life. AI is seen as a powerful tool to enhance such responses. The McGovern Foundation, for example, is working with the American Red Cross to develop AI tools that can do damage assessments. “Immediately after a hurricane, for example, the old and traditional model is they would send out human assessors who would literally walk block by block, taking pictures, and doing a report on every single property,” Dhar said. Using a combination of mapping, visuals, AI, and geospatial intelligence, that timeframe can now be reduced so that financial support can be delivered far more quickly. “AI has the potential to transform a lot of the aspects of humanitarian and development work as we know it.” --— Bernhard Kowatsch, head of the innovation accelerator, World Food Programme Similarly, WFP in collaboration with Google AI has used satellite imaging and machine learning algorithms in an open-source solution called SKAI to get real-time information following a disaster. In the aftermath of floods in Pakistan last year, SKAI scanned over 850,000 households, helping identify the most impacted areas so WFP could provide assistance where it was most needed. “We now have, within 48 hours, results which usually would have taken two weeks,” Kowatsch said. Addressing environmental challenges When it comes to mitigating and adapting to rising temperatures, Dhar explained that large data sets and predictive models can help those working in the climate space better understand the impact of human behavior. For example, the Land and Carbon Lab, a gathering of geospatial AI scientists, is working on building predictive mechanisms to understand how the world's forest cover and carbon reservoirs are changing. “This is an amazing use case for AI and it could have widespread, profound impacts on how we think about climate advocacy and climate transformation,” Dhar said. On a smaller scale, satellite imagery and predictions about local weather can help smallholder farmers make better-informed decisions about what crops to plant and when. Other examples of AI in this space include Digital Democracy’s use of geospatial intelligence and mapping to provide Indigenous Earth defenders with predictive alerts about where and how they might become more effective at dealing with illegal activity in the Amazon; Blue River Technology, which can determine a crop from a weed, allowing for more precise distribution of herbicides; and Microsoft’s collaboration with Indian NGO SEEDS to build maps that identify houses at risk of flooding. “They work with these maps so they can work with the families to help them be better prepared for them,” explained Juan Lavista Ferres, chief scientist and lab director at Microsoft AI for Good Research Lab. Improving global health A range of AI tools are already being implemented across health care, in fields ranging from medical imaging, to robotic surgeries, to automated electronic medical records. By 2030, the World Economic Forum states, AI will be used extensively to detect disease patterns and risks, improve the efficiency of health centers, and decrease patient waiting times. AI in this space isn’t anything new, said Ferres but, in recent years, there’s been an increase in the potential to transform data into comprehensive insights. The majority of data in the health space is stored as unstructured text, such as medical notes, he said. But new AI tools are allowing that to be converted into learnings about particular diseases. For example, over the years India-based development organization Khushi Baby has sent community health workers to villages to educate new mothers about when and how they should seek care. The data collected is now being collated and channeled into a dashboard that provides health officials with insights into communities’ vulnerabilities, so that they can respond with targeted public health interventions. Meza is another digital solution being rolled out by WFP in Kenya. It uses optical character recognition technology to collect photos of patient encounter forms in remote health clinics and digitize handwritten data to eliminate the need for health workers to input this manually. “In the LMIC context with hard-to-reach areas, populations could benefit from AI driven-technologies to provide health care addressing access issues,” Uzma Alam, science policy engagement program lead at the Science for Africa Foundation, said in an email, adding that harnessing such cutting-edge technologies is immense for meeting Africa’s developmental goals in particular. Other examples of AI in global health include Kimetrica’s use of facial recognition to detect malnutrition in children under five within humanitarian emergencies and a chatbot developed by Girl Effect, called ‘Big Sis,’ which answers questions girls in South Africa have about sexual health. At an organizational level In terms of NGO operations, experts said that AI can improve efficiency and save time by doing tasks that might otherwise be manual. “We think that AI can play a fundamental role to make it easy for organizations to be able to work with a lot of data,” Ferres said. Certain programs can, for example, synthesize evaluation reports and analyze satellite imagery, others can write grant applications and draft media releases while some can detect fraud in NGO financials and streamline hiring processes. For cash-strapped NGOs, this could mean more efficient management of resources while also allowing for more effective support for those communities they work with. Viebeg Technologies, for example, uses AI to manage medicine supply chains to ensure health care facilities in Rwanda, Burundi, and the Democratic Republic of Congo are well-stocked; WFP’s Hunger Map uses imaging and AI analysis to give a real-time picture, via an interactive map, of who is hungry, where, and why, allowing them and others to better target their support; and Project Optimus, a supply-chain optimization tool, has also led to savings by identifying — through advanced mathematics that analyzes data points such as population size, transportation routes, and nutritional value of food — the most efficient and cost-effective ways of reaching people. Outside of these categories, interviewees also cited AI as having potential use in tackling disinformation during elections, mapping poverty, and helping those living with a disability communicate. Elsewhere it’s already being used to provide free legal aid to refugees, reunite displaced families, and analyze sustainable fishing practices through underwater robotics. Advice going forward The future potential is unlimited, Alam said, but she warned of both negative and positive impacts. “These negative impacts need to be understood and science policy engagement used to check them in,” she said. “Otherwise, we will be creating a situation of benefiting the few 1% of the global population at the expense of the rest of the world population and our climate and environments.” To ensure its responsible use, Dhar said it’s important for organizations to center people who are at the front line of vulnerability and incorporate their lived experience into the tool design, while Keyzom Ngodup Massally, head of digital programming at U.N. Development Programme’s Chief Digital Office, believes significant stewardship is required. “We believe that stewardship must be more tangible than what we're seeing right now in terms of lots of comments that both marvel and shudder at the advent of AI,” she said, adding that early discourse has centered on how to use AI to radically disrupt life rather than uplift society or support the SDGs. “We do see the imperative now, while AI is still in its nascency … to work with a set of partners and countries to grab a hold of its narrative and mold it in line with ideals such as equity, empowerment, inclusion, and rights,” Massaly said. “If we are able to embed this kind of DNA at the global and local level, we do believe that AI can find its true utility as a force multiplier for the benefit of people and our planet at population scale, rather than just [being] another tool in the long line of industrial innovations.”