🔬 Research Summary by Danny Frank-Siegel, a graduate of Macalester College with a major in philosophy, minor in statistics, and concentration in critical theory.
[Original paper by Dorine Van Norren]
Overview: The near unanimity of Western (meaning European and North American) influences on the design of artificially intelligent algorithms inherently encodes their understanding of the world through Western lenses. This homogenization has led to algorithmic design that philosophically and economically finds itself at odds with cultural philosophies and interests of the Global South. Van Norren attempts to show these issues, and possible solutions, through an analysis of existing Global AI design guidelines using the lens of Ubuntu.
Introduction
When a machine learning algorithm is taught to learn, who teaches it what it means to learn? Translating natural language (languages humans speak) to formal language (computational language) always leaves somewhat of a gap in interpretation, and it skews the biases of the learner to the teacher; in this instance, the learners are highly intelligent algorithms, and the teachers are primarily Western programmers. Van Norren, analyzing both the existing global AI guidelines of various international organizations and the Ubuntu system of thought (both in theory and practice), shows that the difference between Western thought and economics, aligning from Descartes famous’ “I think therefore I am” is at odds with the Ubuntu concept of Ntu, a collective understanding of life, which posits “I am because we are.” Using this alternative lens can be important to new understandings of algorithmic design moving forward, especially as emerging technologies have a larger impact on the African Continent.
Key Insights
What is Ubuntu?
Ubuntu is a system of thought originating with the Bantu people that views people as part of a collective inherently. This means that the individualistic views of Western thought and economics, from an Ubuntu perspective, are incomplete. The individual is defined through relations, to the people around them, to their ancestors, and to the future, rather than just the self; whereas Western Rational philosophy begins with the individual (I think therefore I am), Ubuntu begins with the group (I am because we are). This metaphysical system provides a perspective that interestingly both opposes the capitalist ethos of AI development as it currently exists and simultaneously aligns best with the United Nations Declaration of Human Rights. While the concept of Human Rights forwarded by the United Nations is contentious with Ubuntu based on the definition of individual personhood, the statement of article 27 of the Declaration that “Everyone has the right freely to… share in scientific advancement and its benefits,” suggests that the development of artificial intelligence should not be aimed chiefly at the increase of profits for a few but the net benefit of many. Ubuntu aims to realize that goal, and its practical applications as yet implemented on the global stage make this apparent; the Truth and Reconciliation process initiated in South Africa after the end of apartheid and in Rwanda in the wake of the genocide drew heavily on Ubuntu principles, explicitly and implicitly, to design the process through which the community could begin to heal. The author suggests that a similar process should exist in the design of AI for use on the African Continent, because “If society is or remains organized by the capitalist logic of accumulation of wealth, AI risks going at the detriment of those at the bottom of society (Black 2018, 26). When confronted with economic choices, Ubuntu favors sharing above competition, emulated in the proverb that when one must choose between wealth or the preservation of the life of the other, the latter should be prioritised (Ramose 2005).” (Van Norren, 8).
Guidelines as they exist
There are significant problems in the world today with Global AI infrastructure, such as cybersecurity and AI development/deployment on the African Continent. A cause for concern, as reported by the World Science Forum in 2019, is “the limited number of African researchers and the underrepresentation of African people and data, as well as the lack of full broadband coverage.” (Van Norren, 3). The underrepresentation of African voices in the design of technologies that directly affect African people has highlighted the need for regional frameworks for the development of AI that take into account different cultural values. However, instead of taking a stance on metaphysical or epistemic concerns with the design of AI, COMEST (World Commission on the Ethics of Scientific Knowledge and Technology) has formulated a set of generic AI design principles, which are as of yet insufficient. Fortunately, precedents and examples of the use of Ubuntu in policy shaping exist. While they are only a start in reparations following apartheid in South Africa, the Truth and Reconciliation process is an example of an effective incorporation of Ubuntu into national discourse.
Ntu and the Community: creating a new moral compass
Further, Ubuntu provides an important perspective on what AI ought to be doing when implemented in society: upholding principles that benefit the community. Authors such as David Bamps suggest that AI can help humanity create a “new moral compass.” He argues for the “fyborgisation” of our moral compass, creation of a functional cyborg using both humanity and machine to enforce societal ethics. Both the author and myself disagree, since “AI lacks “Ntu”, the life force. A machine can be intelligent and in relation with people and may alter the community, but the definition of a person includes more than thinking, namely feeling, intuition, animation (the soul dimension) and the capacity to morally grow.” (Van Norren, 8). Van Norren expands upon this idea using the example of robot nannies and care for the elderly. The author warns that “Even if the relationship of caring and robotics is a supplemental at first, the logic of the capitalist society may eventually lead to an increasing role of technology over human aid and (priceless) humane-ness in the relationship… In other words, the ethics of AI have to be considered within the larger context of the dominant ethics of society as a whole.” (Van Norren, 9). However, while Bamps’ fyborgisation does not represent an effective way forward, the idea that AI can be used to help develop a new moral compass should not be discarded; however, AI should still be viewed as a tool to uphold societal well-being rather than a contributing moral agent.
Between the lines
Van Norren’s research establishes an interesting way forward in AI ethics: applying non-Western systems of thought to Western frameworks of algorithmic design. While Ubuntu proposes multiple interesting potential design possibilities, it opens the door for a variety of new questions, namely, are there other non-Western systems of thought that could be more effective for designing global AI systems, and should the concept of a global AI system be abandoned in favor of a series of regional frameworks? While global de-regulation may be ultimately ill advised, consultation on ethical development of AI systems by indigenous groups and the study of alternative ethical systems by programmers so that AI systems are best suited to serve the groups of people they will be directly affecting seems to be a very important way forward in AI development, one that can help shape a “new moral compass” using our most advanced technologies, however that may ultimately look.