CHAPTER 01 / AFRICAN KNOWLEDGE SYSTEMS AND RELATIONAL AI GOVERNANCE
Ethics as Fermentation
“We have all seen ‘responsible AI’ arrive as a late meeting on an already decided calendar. In such moments, ethics is not treated as a condition of legitimacy; it is treated as ceremony.”
Wakanyi Macharia-Hoffman Netherlands
Annette Markham
Netherlands
What We Want Readers to Notice
Too often, ethics enters an AI project after its defining conditions are already settled.
The system has a sponsor. The vendor has made its case. The deadline has acquired the force of inevitability. People affected by the system may be invited to comment, but rarely to govern its pace, purpose, evidence, or limits. Their participation becomes a consultation layered over decisions they had no authority to shape.
When problems emerge, the people nearest the consequences are asked to explain, translate, reassure, and repair. Women, racialized communities, frontline workers, and those carrying relational responsibility are expected to make the system livable without being given power over the conditions that made the harm possible.
We wrote this chapter because ethical principles cannot correct a process whose vessel is already sealed.
Fermentation offered us a different way to think about technological transformation. A batch does not become nourishing simply because it changes. Its outcome depends on the conditions under which it is made: its heat, vessel, lineage, starter culture, and capacity for rest.
Ubuntu and Ma’at deepen that argument. They refuse to separate individual agency from reciprocal responsibility or technical truth from justice. They ask not only whether a system works, but what relationships it sustains, whose knowledge governs it, what costs it displaces, and who retains the authority to slow, reopen, redesign, or refuse it.
We developed ethics-as-fermentation to move governance upstream—before speed becomes destiny, enclosure becomes maturity, and repair becomes someone else’s unpaid responsibility.
By the Authors
Wakanyi Macharia-Hoffman | Annette Markham, Ph.D.
Wakanyi is an African Indigenous scholar of Ubuntu philosophy whose work advances the integration of Ubuntu ethics into AI design and governance. She is the lead researcher for Sustainable African AI Design at the Inclusive AI Lab, part of Utrecht University’s Centre for Global Challenges.
Annette is Professor of Media Literacies and Public Engagement at Utrecht University and founder of the Futures+ Literacies + Methods Lab. An internationally recognized digital ethnographer and methodologist, she examines how AI, datafication, and digital media reshape identity, everyday life, and social practice.
FEATURED RESOURCE
Tending the Batch
A practical toolkit for governing the conditions of AI-making
The Tending the Batch Toolkit helps leaders examine an AI project before its assumptions, incentives, and harms become embedded in infrastructure.
Rather than beginning with a general question such as “Is this AI ethical?”, the toolkit examines five conditions that shape what the system can become.
Fermentation is not used here as a decorative metaphor. Ubuntu and Ma’at are situated intellectual and governance traditions—not interchangeable shorthand for a single or universal “African ethic.” The toolkit should be used with attribution, context, and respect for the authority of the communities whose knowledge informs it.
FOR
AI-governance teams, institutional leaders, public agencies, NGOs, procurement groups, designers, educators, community partners, data professionals, and responsible-innovation programs
TIME
60–90 minutes for a facilitated review
FORMAT
Downloadable scorecard, governance ritual, optimization questions, and declaration template
Designed to be tested and adapted for your setting—not followed as a fixed prescription.
Why This Matters
NOTICE
Recognize the conditions already shaping an AI project before formal ethics review begins.
Notice when urgency is silencing affected communities, contracts are sealing the vessel, inherited categories are being laundered into technical neutrality, community knowledge is being treated as commentary, or no one has authority to stop the work.
DECIDE
Turn ethical concern into a consequential governance decision.
Use the Fermentation Scorecard and Ulimisana questions to determine whether the project should proceed, proceed under conditions, pause, be redesigned, or be refused—and document why.
SUSTAIN
Build accountability into the system’s lifecycle rather than relying on individual vigilance.
Establish audit rights, scope limits, consent renewal, review dates, contestation pathways, repair obligations, stopping rules, and named authority to call for pause or refusal after deployment.
Technology governance is also an evaluative question: whose knowledge counts, whose context matters, and who has authority to name harm, success, and repair. By centering African, Indigenous, feminist, and justice-centered ways of knowing, this volume challenges the epistemic violence of technological defaults and offers practical pathways for more accountable governance.”
PROFESSOR BAGELE CHILISA
Postgraduate Research and Evaluation Programme University of Botswana Author of Indigenous Research Methodologies
“
ABOUT THIS CHAPTER
AI ethics is commonly presented as a collection of principles, review practices, or safeguards added to a system. This chapter argues that such approaches often begin too late.
By the time an ethics committee, affected community, or relationally positioned leader enters the room, the system’s pace, vendor, data lineage, institutional purpose, and commercial incentives may already be treated as settled. Ethics then becomes ceremonial: it can comment on the project but cannot alter the conditions producing its risks.
Ethics as Fermentation reframes AI as a governed transformation. Like a living batch, an AI system becomes nourishing or harmful depending on the conditions under which it is developed and sustained.
The chapter draws on Ubuntu and Ma’at to bind agency to reciprocal responsibility and truth to justice. Moral AI Agency identifies the capacity to act before harm is locked into infrastructure. Ulimisana gives those commitments operational force by treating documentation, bounded purpose, consent, contestability, dignity, and refusal as constraints on what optimization may pursue.
The practical contribution is a governance architecture in which pause, redesign, and refusal become evidence of technical and leadership competence—not signs of resistance to innovation.
SUGGESTED USES
AI project inception, readiness, renewal, and reconsideration reviews.
Procurement, vendor evaluation, and public-sector technology assessment.
Responsible-AI, data-governance, and model-risk workshops.
NGO, community, and participatory technology-planning processes.
Leadership, executive education, and responsible-innovation programs.
Courses in feminist technology studies, science and technology studies, Indigenous and decolonial methods, and AI governance.