AI for development work
The debate around AI is now fuelled by the discussions around the positive and negative impact AI technologies can have on our society. More and more initiatives that foster the thinking of leveraging AI for the SDGs are popping up, exploring how AI might support development work to achieve more impact (see, for example, AIxSDGs or AI for Good).
These initiatives are important to bring innovative ideas and approaches to the development and humanitarian field. What becomes evident though is that for development work, innovative approaches, including the use of AI, often focus on implementation approaches and interventions in the field. What is often still lacking is thinking about how innovative approaches can strengthen the capacities of development organisations themselves to deliver their work more efficiently and effectively.
Ever since starting our work, at Strhive, in the field of international development cooperation, we’ve come across this sentence many times: “if only we knew what we know”. And not to say this is a sentiment restricted to development organisations, it can also be found across large, well-established corporates. Nevertheless, for development organisations there is a more profound implication. Considering all the knowledge development and humanitarian organisations have already and continue to produce, imagine you could simply ask a tool: "what do we know about what works in this context on this topic for sustainable development?"
Something to think about
One of the problems with knowledge and learning in the field of development is that while a lot of data, quantitative and qualitative, is collected through all sorts of studies and evaluation, it is not used in future programming.
“Actively ‘not learning from experience’ is as much a part of organizational processes as learning from experience” (Hulme, 1989, p. 1).
There are several reasons for that. One key reason is that the large amounts of qualitive data and experienced-based knowledge is hard to digest. Workload is already immensely high in the development field and there is often no time or capacity to digest hundreds of pages of reports to draw strategic conclusions from. Another reason is that the outcomes of, for example, evaluation studies often do not reach the audiences within organisations in a way that they could be reused and applied in future programming (Carlsson, 2000). Lastly, the decision-making power over which lessons and knowledge count does not lie with the people who might know best and political agendas and power relations can influence how lessons and knowledge are used (Hulme, 1989).
These dynamics inherent to international development work are important to consider when discussing how knowledge can be used. Specifically as reaching the SDGs requires a coordinated approach involving many stakeholder:
“It is coordinated because government, industry, academia, and society must work together to reach the SDGs (Goralski and Keong Tan 2023). Actions also need to be coordinated at different levels. For example, different stakeholders should complement each other’s actions, for example, industry can provide innovative technology.” (Mazzi and Floridi, 2023, p. 12).
AI can lend a helping hand in digesting this wealth of knowledge, helping us understand what is out there, what works and what does not. But considering the discussions around regulating AI and its biases and the discussion in international development around “whose knowledge counts” we must be cautions. The discussions around the application of AI for development have to be embedded in the conversation around power dynamics and decolonisation in international development.
We have to ask ourselves first about the role of politics and power relations before employing AI for capacity building for development organisations and keep in mind that:
“The goal of zero poverty is threatened by the imperfect design and implementation of decision-making algorithms that have displayed evidence of bias, lack ethical governance, and limit transparency on the basis of their decisions, causing unfair outcomes and amplifying unequal access to finance (Truby 2020).” (Mazzi and Floridi, 2023, p. 11).