Opinion: AI in global development is more than just a set of tools

With the aid of artificial intelligence, international development professionals have succeeded in introducing multiple tools that offer practical solutions to challenging problems. Beyond that, the AI ecosystem itself can contribute to creating an enabling environment for sustainable development by engendering a need for a trained labor workforce. The good and the bad news Combining available data, local expertise, and machine learning technology has presented new opportunities to approach economic development issues and improve decision-making. The U.S. Agency for International Development for example supports integrating AI in their funded programs. Its inventory of AI use cases lists 15 activities for 2023. In health, for example, USAID used AI to optimize COVID-19 vaccine allocation, to predict bed occupancy at hospitals, and to forecast tuberculosis drug quantities. Other examples include AI applications that range from social media listening to mapping urban vulnerability to detecting rhino horns in travelers’ luggage. These initiatives showcase the positive impact of AI tools on global development efforts. While the focus is on the benefits of such applications, AI is also capable of introducing harm. USAID’s report on “Making AI Work for International Development” highlights areas where models can go wrong — which are mainly related to accuracy in training data or in the algorithm. From my experience in international development, depending on the field and the geographic area, data can be inaccurate (due to data entry errors), incomplete (due to lack of data entry), or simply not available (does not exist). Other related challenges manifest in the unwillingness to share data for reasons of privacy, private ownership, security, or authority, to list a few. The algorithm, on the other hand, might be implemented by technology developers not familiar with individual country contexts. This has the potential of injecting biases or personal influences which negatively affect the accuracy of the output. USAID’s report also lists a few risk factors, including excessive trust. In literature, the topic of AI and trust is heavily researched and mentions overreliance on AI outcomes. The adoption of AI tools has an inverse relation with people’s skill levels. According to research, the more experienced they are, the less likely they are to accept AI recommendations in their decision-making process, and vice versa. In the case of low-income countries, the introduction of AI applications to solve strategic or critical problems has to consider — and mitigate — the harms linked to lack of accuracy and overreliance on AI. This means we need to strengthen the data and human resources infrastructure to sustainably support both human and digital resources. The AI ecosystem as a job mine While many fear they might lose their jobs to AI, this is not a major human resources concern in the development world. In fact, the AI ecosystem could be a job mine for certain economies once they are equipped with the proper skills. The USAID Artificial Intelligence Action Plan suggests approaches to “strengthen ecosystems to support responsible use of AI” and elaborates on how USAID programming can foster AI readiness in the labor market. Preparing a workforce capable of continuously using, maintaining, and upgrading any AI tools that have been delivered as part of donor programming is essential to program sustainability. Too many initiatives have failed to survive after donor support ended. When it comes to job readiness, however, most programs focus on technical and soft skills training as a means for preparing program participants for jobs. In my experience, this supply-side approach did not achieve satisfactory job placement results and was outperformed by a demand-based design qualifying unemployed people for available vacancies in partnership with the private sector. In the case of AI as an industry, demand will include skills supporting the whole ecosystem such as data-related functions, algorithm development and coding, maintenance and sustainability of AI tools, infrastructure, integration, ethics, etc. Training programs and/or university curricula must be established in partnership with the industry to reflect their systems and hiring needs. AI niches In a country like Jordan, where the Country Development Cooperation Strategy integrates women and youth across development objectives, a sustainable AI ecosystem ensures gains at multiple levels of the results framework. While cultural stigmas may exist toward certain occupations in Jordan, the ICT sector has a competitive advantage in attracting more women and youth into the workforce. Women comprise 50% of Jordanian university ICT students and 30% of ICT employees — which is about double the national average for women's participation in the labor market. Another niche of AI and ICT applications is that they cut across multiple sectors. Skilled personnel can thereby establish careers in health, finance, democracy and governance, agriculture, education, and others. Moreover, the demand for AI expertise globally signifies that such a trained labor force will not be restricted to a small economy and can succeed in exporting the knowledge, thus supporting other countries’ efforts to adopt this technology. AI is not just a set of useful tools. AI is a disruptive technology ecosystem and enabling environment, with the potential to create job opportunities, achieve development objectives, and work toward achieving the Sustainable Development Goals.