AI for Development is a Malapropism
Abstract
Artificial Intelligence (AI) has increasingly been included in the range of enabling technologies and tools that support and streamline the process of Development. This led to the promotion of domains such as AI for Good, AI for the Poor or AI for Development (AI4D). But, doing so, it is implicitly assumed that AI complies with the distinctive features of ICT4D such as affordability, relevance, openness and ownership. In this paper, it is shown that AI does not fulfil most of these characteristics and that the term AI4D does not seem to be grounded. To a certain extent, the use of this acronym is misleading and against the fundamentals of the Technology for Development. Decision makers in developing countries shall ponder all factors related to the use of AI and embrace a wider perspective in elaborating and sustaining their national digital transformation strategies.
Keywords
Download Options
Introduction
Actors involved in ICT4D (Information and Communication Technology for Development) can be naturally tempted to include Artificial Intelligence (AI) in the range of enabling technologies and tools that may support and boost the Development process. This may be a legitimate expectation. Yet, promoting AI for Development (AI4D) raises numerous problems since it assumes that AI complies with the very specific features of the ICT4D context, including, among others, affordability, relevance, openness, and ownership. Unfortunately, today’s AI fulfils none of these, and the enthusiastic posture that AI4D is a natural part of ICT4D is not grounded and, to a certain extent, does not serve the purpose of Technology for Development.
With the purposes and goals of today’s AI technology, it is difficult to elaborate a research proposal where an AI application/tool is used to primarily serve the purpose of Development, without explicitly admitting (and contributing) to some amalgamations and/or maline abstractions, including: technology misconception, mixing the application domain and the technology type, stating/defending general human rights, social and societal values and principles, and how technology could support these, regardless of whether this technology is ICT4D or not.
Considering the current race that takes place in developing countries and aims at elaborating digital transformation strategies using AI as a core technology (Government of India, 2018), it is timely and useful to (re)put things in the appropriate context, and define/understand their meaning, contrast them, and conclude as to how fit these are within the Development context and constraints. It is not healthy to adopt a technology just because it is new and/or because vendors pretend it can solve all our problems. Rather, we believe that policymakers (in developing countries, particularly) should concentrate on the actual needs (of the country and the population) and to define them independently of any specific technology nature or type. Then, they could prioritize ICT4D (Model, Techniques and Tools), as the main enabler for the digital transformation process of their countries.
Conclusion
During the past decade, through slogans such as AI for Good, AI for the Poor and AI for Development (AI4D), AI has been strongly promoted to support and accelerate Development. This ‘AI frenzy’ may potentially overshadow the continuous efforts needed to sustain ICT4D in developing countries, at least in the mind of citizens and of policy and decision makers.
In this paper, we shed some light on an important aspect that seems to be forgotten in the current AI4D promotion: the principle that using any technology in ICT4D should first promote/support Development and contribute to good governance in general. We highlighted the contrast between the nature and purpose of AI, and its purported revolutionary use as an ICT4D technology. Although, at first glance, AI4D seems comparable to ICT4D, we showed that the analogy does not stand.
Artificial Intelligence, as a field of knowledge, is undoubtedly interesting and exciting! It has great potential, but it does not particularly fit in the context of Development and is not inherently ICT4D Compatible. ICT4D is different from ICT, and from AI, taken alone. The difference does not only encompass the context, specifications and methodologies that are strongly linked to Development, but they shall fulfil specific requirements derived from the development context including, among others, affordability, openness, relevance, ownership, understandability and explainability. We showed that current AI applications do not comply with these requirements. As an alternative, we propose a vision that fosters ICT4D thanks to the wise use of relevant AI applications, in the same way as any other enabling technology. We believe that in the context of development, this vision is a better fit and more reflective of what (and how) to conceptualize the benefits of the AI technology, without compromising the ICT4D fundamentals. We are currently working on a methodological framework to accompany policy and decision makers who would like to adopt such a vision.