Anekantavada

Global narratives around AI center the visions and ideas of those in positions of power. Data often represent dominant parts of the world without accounting for diversity and are filled with bias. The models trained on them are used by and on people who were not represented in the training data. These visions and ideas are also determining the narratives around AI, sometimes pushing AI like it is a silver bullet, and sometimes turning to fear-mongering. Techno-optimists are also moving along, suggesting a “be there or be square” narrative, pushing for people to learn to use AI or risk being left behind. On the other end of the spectrum, those turning to use AI by leaning on the promises it offers to help level several different playing fields are forced to reckon with the dilemma of using the technologies for its environmental harm, bias, and extractive nature. The reality remains that AI systems are built, undergirded, and pushed forward through partial perspectives that do not account for diversity in its entirety. Anekantavada offers us an opportunity to expand the scope of perspectives in the AI lifecycle.

Before we proceed, we want to state that Anekantavada is a living philosophical and epistemological doctrine within the Jain tradition, which is among the oldest continuously practiced religious and philosophical traditions in the world, with communities primarily in India and a global diaspora. Jainism is a minority tradition that has often been under-represented in global philosophical discourse despite its extraordinary intellectual contributions to epistemology, logic, ethics, and environmental thought. Anekantavada is closely related to two companion doctrines: Syadvada, the theory of conditioned predication, which holds that statements can only be conditionally true; and Nayavada, the theory of partial perspectives, which holds that any single viewpoint captures only one aspect of reality. Together they form a comprehensive epistemological framework that contemporary scholars have applied to logic, conflict resolution, and interreligious dialogue. We welcome correction and deeper engagement from Jain scholars and communities who carry this tradition.

Understanding Anekantavada

Anekantavada is conveyed best through a fable. A group of blind men are each asked to describe an elephant. One touches the trunk and calls it a snake. One touches the leg and calls it a tree. One touches the tail and calls it a rope. One touches the ear and calls it a fan. Each is entirely correct about the part they have touched, but each is entirely wrong about the whole in terms of what they identified. Anekantavada, the doctrine of many-sidedness, teaches us that reality is multi-dimensional and that no single viewpoint can capture it completely. It is the understanding that to act on one perspective as though it were the whole truth is a subtle form of violence that goes against truth, other people, and the irreducible complexity of the world.

The word “Anekantavada” is composed of three Sanskrit elements: 'a-' (non), 'ekanta' (one-sided, absolute), and 'vada' (doctrine, teaching). It is the doctrine of non-one-sidedness: the teaching that reality is many-sided and that no single proposition can exhaust its full nature. Anekantavada does not say that all perspectives are equally valid, or that they all constitute an absolute truth. Instead, it reminds us that any single perspective, however accurate within its scope, captures only part of what is true. The task is not to abandon judgment but to hold it with epistemic humility, while remaining open to what other perspectives reveal. Dogmatic one-sidedness is considered a subtle form of violence toward both truth and the people whose perspectives are dismissed.

What Anekantavada unsettles in AI governance

Anekantavada makes a powerful contribution to AI governance at the level of epistemic design. AI systems are built on training data that samples the world from specific vantage points and learns, from that sample, draws patterns from it, and then applies it universally. The system that has learned from text produced primarily by English-speaking, Western, high-income populations treats these patterns as the default description of how language, reasoning, and the world work. Communities whose perspectives are absent from or marginal in the training data find that the system describes them inaccurately, evaluates them unfairly, and excludes them from the world that the system models as normal. Governance informed by Anekantavada would considers this epistemic violence.

Applied practically, Anekantavada would require AI governance to ask, about any AI system: whose perspectives shaped this system’s understanding of the world, and whose did not? What does the system see clearly, and what is it constitutively unable to see because of the angle from which it was built? What are the conditions under which its judgments are reliable, and the conditions under which they will inevitably be wrong? These are questions that current AI evaluation frameworks approximate through fairness testing and bias auditing, but do not fully reach, because fairness testing still assumes that there is a correct answer against which performance is measured. Anekantavada challenges that assumption: if reality is many-sided, the correct answer to many of the questions AI systems are asked does not exist — it depends on the standpoint from which the question is posed.

Anekantavada also carries a specific governance recommendation: governance processes should be structured to make as many relevant perspectives visible as possible before any binding decision is made. This is not a mere call for stakeholder inclusion, which can be satisfied by inviting a diverse set of people to comment on a decision that has already been substantially made. It is a call for structural epistemic pluralism, where governance processes are designed so that different perspectives can genuinely change the analysis, and not just the participant list. A governance framework shaped by Anekantavada would not produce a single correct AI ethics standard and apply it universally. It would produce conditional standards that may be true from a particular standpoint, under particular conditions, and for specific communities, subject to revision as more and new perspectives are incorporated.

References

Matilal, B.K. (1981). The Central Philosophy of Jainism: Anekanta-Vada. L.D. Institute of Indology.

Gupta, S. & Rajput, D. (2025). Anekantavada: The Jain Doctrine of Many-Sidedness and Its Relevance in the Modern World. Praxis International Journal of Social Science and Literature. 7. 62-66. 10.51879/PIJSSL/070709.

Long, J.D. (2009). Jainism: An Introduction. I.B. Tauris.

Chapple, C.K., & Donahue, J. (eds.) (2003). The Jain Concept of Nonviolence. SUNY Press.

This article draws from Jain philosophical traditions that have sustained the principle of Anekantavada across millennia. We are grateful to the scholars, practitioners, and communities who carry these traditions and have made their wisdom available for wider engagement. We invite readers to support organizations working on data justice and digital commons, which carry the Anekantavada spirit into contemporary governance contexts.