• logo linkedin
  • logo email
intelligence artificielle ordinateur programme numerique AFD dévleoppement innovation odd
Public development banks are taking an interest in AI tools to improve the impact of their funding, starting with AFD, which presented its SDG Prospector in June. For Régis Marodon, Sustainable Finance Advisor at AFD, "artificial intelligence's foray into development is a potential watershed".

A growing number of institutions and development stakeholders are beginning to examine and deploy artificial intelligence (AI) in relation to the United Nations' Sustainable Development Goals (SDGs). Roundtable talks will focus on the use of data analysis and AI to enhance the impact of development projects at the annual public development banks meeting, the Finance in Common summit, held from September 4 to 6 in Cartagena, Colombia

This is only the beginning. "Artificial intelligence's foray into development is a potential watershed," says Régis Marodon, economist and Sustainable Finance Advisor at AFD. "It's a new frontier, something that could radically change the way we work, like the arrival of computers in our offices. We'll be able to focus our expertise on the impacts and relationships with our partners.”

With the new methods made possible by recent advances in this field, it is now possible to analyze large quantities of data on development funding in an instantaneous, secure, standardized, reliable, and verifiable way.

"These systems were originally developed by Facebook and Instagram to contextualize texts, understand what users like or dislike, and generate targeted advertising. Why shouldn't they be used to work on the complex issues of sustainable development?" asked Régis Marodon.

SDG Prospector: a new tool

Sustainable Development Goals are still often monitored manually, using reporting data. "This approach naturally has its limitations: monitoring is not systematic, can be subjective, and the methodology used is not necessarily the same from one institution to another," pointed out Adeline Laulanié, data officer at AFD.

AFD Group has just designed a new tool, the SDG Prospector, to assess references to SDGs in any type of document. Introduced in June 2023, it is based on a language model developed by Facebook, which enables it not only to identify keywords, but also to contextualize sentences and analyze documents submitted to it in greater detail.

See also: "ChatGPT": AFD Group puts artificial intelligence to work for the Sustainable Development Goals

"We've fed it with nearly 9,000 texts from the UN, governments, NGOs, project documents, which sum up the world of SDGs. It now understands when a paragraph does or doesn't mention it," says Jean-Baptiste Jacouton, researcher in sustainable finance at AFD, who has been involved in the project from the outset.

The SDG Prospector can map portfolios and report on a stakeholder's activity based on various criteria. Does AFD support the fight against poverty more through its grants or loans? In which countries does AFD contribute most to SDG 6 on water and sanitation? Answers are provided in just a few seconds. 

Certain SDGs are overlooked

The analyses already conducted have revealed a first lesson: cross-cutting SDGs such as gender equality, reducing inequality and eradicating poverty still represent a minor part of the discourse of public development banks, contrary to what one might think.

The algorithm can also account for historical trends: find out, for example, whether an institution funds more or fewer water projects in drought-affected regions, check whether it supports as many climate programs as it claims, and when and where they were carried out. It's a way of ensuring that strategic objectives are translated into tangible results on the ground.

See also: Digital solutions and innovation at AFD Group

"Its job is to look for needles in the haystack," says Régis Marodon. "Our aim is to provide an analysis tool that will become almost as essential as Excel on a computer."

With a success rate often in excess of 90%, the SDG Prospector is already a valuable asset to support human analysis. It is not yet in operational use - it's only a prototype - but everyone is invited to test it on the dedicated website.

The objective now is to continue the research and teach the algorithm to identify the 169 targets linked to the 17 SDGs. "It will then be able to work on the interactions between targets. For example, a sanitation project has an impact on river pollution control, as well as on ocean pollution control and indirectly on small-scale fishing. We'll be able to trace the causal chains in a true sustainable development approach," says Marodon. 

International research group

Several institutions such as the OECD (SDG Tracker), the UN (LinkedSDG), and the European Union (SDG Mapper) have developed their own algorithms for mapping the SDGs, which involve counting word frequencies. The drawback is these traditional approaches do not allow documents to be interpreted in context.

An international research group has been set up to promote exchanges on these AI and sustainable development issues. It includes some thirty specialists from the United Nations, the private sector, and public development banks. The rationale is collaborative: trying to share the best elements from everyone.

In the future, it will be possible to feed these programs with immense volumes of data and ask them to identify unnoticed causal links. This is one of big data's long-standing promises, dating back to the early 2010s. By having access to project impact data, for example, it would be possible to learn new lessons systematically and avoid repeating mistakes. 

"The value of AI is obvious," says Marodon. "Development banks have a greater responsibility in aligning their funding with the Sustainable Development Goals, and consequently a greater responsibility in promoting anything that can contribute to them."