From Rationality to Relationality

Ubuntu as an Ethical & Human Rights Framework for Artificial Intelligence Governance

Published

Carr Center for Human Rights Policy, Harvard Kennedy School — Discussion Paper 2020-009, July 2020

Written as

Technology and Human Rights Fellow, Carr Center

Length

28 pages · 147 footnotes

The argument

The discriminatory effects of automated decision making systems are not accidents of implementation. They are tied to the philosophical origins of artificial intelligence itself — to the Western world’s early conception of what it means to be a person.

That conception is rationality. Aristotle’s rational animal, Descartes’ I think therefore I am, Kant grounding moral autonomy entirely in reason. The same lineage runs directly into computing: Lull’s logic machine, Leibniz’s calculus ratiocinator, Boole, Frege, and Turing’s thinking machine. Artificial intelligence inherited a definition of personhood, not merely a set of techniques.

But that definition has always been marked by contradiction and exclusion. Personhood as rationality was never applied equally: it supplied the moral license for colonialism and enslavement, and it shaped the economic structures (capitalism’s free markets built on colonialism’s forced markets), political structures (modernity’s individualism imposed through coloniality), and social hierarchies that now shape AI’s data, creation and use. It is unsurprising that the industry reproduces these dehumanizations.

History reveals modernity’s dehumanization in that irrational measures are often taken in the pursuit of rationality.

There is also a formal argument. Gödel’s incompleteness theorems show a consistent formal system cannot prove all truths within itself — another system is needed. If logic is a consistent formal system, and logic is the basis of human reasoning, then no individual is self-complete while their own system of reasoning is incomplete. Rationality is therefore not an individual product but a result of interconnected systems: a consistent formal system is complete through other consistent formal systems. Substitute personhood for formal system and the aphorism follows — a person is a person through other persons. Personhood is fundamentally relational.

Ubuntu as relational personhood

The Sub-Saharan African philosophy of ubuntu — more precisely ubu-Ntu, where Ntu is the root of person, people and personhood — holds that one is not complete in oneself but relies on others for completion. Ethics is the measure of one’s relationality: to others, to the environment, and to the interdependent whole. The paper reconstructs this from five Nguni proverbs, each yielding a principle applicable to human rights law and AI governance, including Umu-Ntu akalahlwa (no one is beyond redemption — the logic behind South Africa’s Truth and Reconciliation Commission) and Inkosi yinkosi ngaba-Ntu (a leader derives power from the consent of the governed).

Ubuntu may also supply the philosophical justification for the third generation of human rights — the rights to solidarity and community adopted by the Organization for African Unity in 1981. The African Charter on Human and People’s Rights remains the only internationally binding document containing all three generations.

Five critiques of artificial intelligence

Any harmful automated decision making system suffers from one or more of these five defects, each a violation of ubuntu’s relational personhood:

1. Exclusion of marginalized communities from the design, development, decision making and funding of these systems.
2. Biases in procedure and data — both in the selection of features and in the data itself, embedded through value-laden choices and then closed to public scrutiny.
3. Failure to recognize the interconnectedness of society — the misguided belief in technological neutrality, when those with social protections can offset harms that others cannot.
4. Commodification of our digital selves — reduction of a person to commodifiable metrics. Machine learning is digital divination, but without the cooperation and consent that holistic divination requires.
5. Centralization of data and resources — ceding control to a few corporations, countries and continents without public accountability.

What it proposes

The paper closes with technical and policy recommendations: decentralized training where user data never leaves the device; differential privacy as a mathematical guarantee; an inverted data model where the platform sends a “recipe” and the browser assembles the “ingredients” locally without the data ever returning; offline peer-to-peer networks over Bluetooth, Wi-Fi and ultrasonic transport. On the policy side: public impact assessment bodies, legislation giving communities control over their data as licensable property, and rules preventing large platforms from foreclosing local markets through subsidized data bundles.

Ethics is not missing in technology. Our ethics and values are always present in the creation and use of technology. The technology society creates and chooses not to create is a window into the ethics and values of the powerful.

Why this work and government work are the same project

A reasonable question: how does one critique the Western philosophical tradition while working inside Western institutions? The paper answers it directly, and the answer is not a compromise.

The argument is not that rationality is wrong. It is that rationality is incomplete — and that this is a formal result, not a grievance. Gödel showed that a consistent formal system cannot complete itself; it requires another system to do so. Applied to personhood, that is precisely ubuntu’s claim: a person is a person through other persons. A tradition that insists on its own self-sufficiency is making a claim mathematics already refused.

So the work follows from the argument rather than contradicting it. Institutions complete each other. My antitrust analysis for the House Judiciary Subcommittee made a structural argument — that concentration across the AI stack harms those with the least power to absorb it — in the vocabulary of American competition law and the Subcommittee’s own 2020 findings. It reaches a conclusion this paper predicts, through an institution’s own reasoning. That is what it looks like to take a tradition seriously enough to argue inside it.

The paper closes by calling ubuntu one of Africa’s greatest gifts to the world. A gift is offered to someone. It is not a wall.

Contents

1. Introduction
2. Individualism: the irrational personhood
3. Ubuntu as Relational Personhood
4. Implications of Relational Personhood
5. Data Colonialism and Surveillance Capitalism as Attacks on Relational Personhood
6. Five Core Critiques of Artificial Intelligence
7. Ethics and Technology
8. Technology and Policy using Ubuntu
9. Conclusion

Statements and views expressed in the paper are solely those of the author and do not imply endorsement by Harvard University, the Harvard Kennedy School, or the Carr Center for Human Rights Policy.