The Traditions

TAM-INS.02 · The Insufficient · The Approximate Mind

Priya has not slept well in months. She is forty-two, and she works land that belongs to her husband’s family in Satara district, about three hours southeast of Pune. She also raises three children, walks to a water source that has moved further away twice in five years as the local well’s output has declined, manages a household in which she is the only fully functional adult, and carries in her body the accumulated record of all of this.

She walked into a primary health center last March. The AI triage system processed her. It asked her symptoms. It cross-referenced her bloodwork, her age, her listed occupation. It recommended iron supplements and rest.

The previous essay argued that a system capable of questioning its own categories, a skeptic, would have caught the insufficiency. This essay asks a harder question. There is not one way to see what the system missed. There are many. And each way of seeing catches something the others accept.

The Problem with a Single Skeptic

#

The first essay named the danger: a skeptic trained on any single tradition will doubt what that tradition taught it to doubt and accept what that tradition did not flag. A single skeptic is a single shape of doubt, and the shape has blind spots as dangerous as the ones it was designed to catch.

The solution is not one skeptic but several. Not seven worldviews tokenistically represented. Seven operations, each derived from a philosophical tradition the AI development ecosystem was not built to see. Each operation is a specific test applied to a specification before the specification becomes operational. Each catches what the others miss.

The analytic philosophical tradition that dominates AI ethics already provides logical rigor and formal clarity. That tradition is the water the entire system swims in. You do not need a skeptic trained in the tradition that built the system. You need skeptics from the traditions the system was not built to see.

Here is what each one catches when it encounters Priya.

Permanent Suspension

#

Pyrrho of Elis, third century BCE, practiced a form of skepticism that did not resolve. Where Descartes doubted in order to find certainty, Pyrrho suspended judgment as a permanent condition. The Pyrrhonist never arrives. The doubt is not instrumental. It is the practice itself.

The operation: flag every category in the specification that is treated as a natural kind but has not been independently established as one.

Applied to Priya: “Patient” is a classification decision, not a fact about her. “Agricultural worker” is a category imposed by an intake form, not a description of her life. “Fatigue” is a word that the triage system treats as a medical entity. It could equally be treated as a biographical fact, an economic consequence, a structural condition, or an honest report from a body that has been asked to do more than any body should.

The Pyrrhonist does not say these categories are wrong. It says they are provisional. Each one was treated as a starting point. The Pyrrhonist insists they are conclusions, reached without argument, accepted without examination.

This is the spine of the architecture. It provides the posture. The other six provide the specific tests.

Anti-Reification

#

Nagarjuna, the second-century Buddhist philosopher, argued for the emptiness of inherent existence. Not nihilism. The precise claim that no category possesses independent, self-sustaining reality. Everything is dependently originated. A thing exists only in relation to other things, never on its own terms.

The operation: identify every variable in the specification that is being treated as a stable, context-independent entity. Flag the moment an abstraction starts being treated as a thing.

Applied to Priya: “Fatigue” is not a thing. It is a word that the system treats as an entity with properties, a symptom to be matched against a catalog of conditions that produce it. But fatigue-as-Priya-experiences-it is not the same entity as fatigue-as-the-clinical-literature-defines-it. Her fatigue is not separable from the walk, the water, the household, the monsoon, the years. The system reified her exhaustion into a medical category, and that act of reification determined what interventions were possible before any intervention was considered.

The reification is invisible because it happens before the analysis begins. By the time the system generates a differential, it has already decided what kind of thing it is looking at. That decision was never examined. It was built into the intake form.

“Crop yield per hectare” reifies both the crop and the hectare. “Patient satisfaction score” reifies satisfaction. “Student performance” reifies performance. In each case, a relationship is converted into a property, and the conversion is treated as measurement rather than construction. Nagarjuna’s operation catches this conversion every time it occurs.

Relational Ontology

#

Ubuntu, the Southern African philosophical principle, is often translated as “I am because we are.” The formulation sounds warm. Its intellectual content is radical. It says the unit is the relationship, not the individual. A person is not a self-contained entity with properties. A person is a node in a web of mutual constitution.

The operation: for every specification that treats a person as an isolable entity, identify the relational web the specification has severed. Map the relationships the model cannot see because it has already decided the unit is the individual.

Applied to Priya: her health is not a property of her body. It is a property of her household, her water access, her marriage, her children’s school situation, the monsoon pattern, the local economy, the caste structure that determines her access to certain kinds of help, the gender norms that assign her the labor nobody else will do. Treating her as an isolable patient with an isolable symptom is a category error that no amount of better data or better algorithms can correct. The error is in the unit, not in the analysis of the unit.

The triage system processed one person. Reality contains a web. The system’s output, iron supplements and rest, is a recommendation for the node. It has no recommendation for the web, because the web is not in its ontology.

Situated Knowledge

#

Sandra Harding, Nancy Hartsock, and Patricia Hill Collins developed standpoint theory across several decades of feminist epistemology. The core claim: knowledge is situated. There is no view from nowhere. Every claim to objectivity is a claim made from a particular position, and the position shapes what can be seen.

The operation: for every specification, ask from whose position this looks like the right question. From whose position does the proposed solution look adequate?

Applied to Priya: “Iron supplements and rest” looks like an adequate response from the position of the clinician with thirty patients waiting. From the position of the system designer in Bangalore who trained the model on clinical guidelines. From the position of the grant agency that funded the triage system and measured its success by throughput and diagnostic accuracy.

From Priya’s position, rest does not exist. It is not a treatment option. It is not available to her. Her life does not contain it. Recommending rest reveals that the system has no model of her actual constraints. It generated a recommendation from a position that assumed rest was possible, and that assumption was never examined because the people who designed the system live in a world where rest is possible.

The system performed competently. It also performed from a standpoint that was invisible to itself. That is the definition of privilege operating as objectivity.

Consequential Verification

#

Charles Sanders Peirce, William James, and John Dewey built American pragmatism on a simple challenge: truth is what survives inquiry. Not “is this true in the abstract” but “what happens when you act on it?”

The operation: trace every recommendation, classification, or optimization to its six-month consequence. If acting on the recommendation does not resolve the condition, the recommendation was insufficient regardless of its internal validity.

Applied to Priya: if she takes the iron supplements and continues her life exactly as it is, what happens in six months? She comes back. Same symptoms or worse. The iron may have marginally improved her hemoglobin. The fatigue persists because the fatigue was never an iron problem. The joint pain persists because the joints are still carrying the same load.

The recommendation passes a clinical adequacy standard. It fails a consequential one. The pragmatist does not care whether the recommendation was well-reasoned. The pragmatist cares whether it works. And “works” means: does the person’s situation improve? If not, the reasoning was insufficient, no matter how internally consistent.

Non-Transferability

#

This is not one tradition but a shared structural feature across multiple Indigenous knowledge systems: Maori, Aboriginal Australian, First Nations, and others. Knowledge is not separable from the knower, the place, the relationship, the responsibility that comes with knowing.

The operation: for every specification that assumes knowledge is portable across contexts, identify the contextual dependencies the portability assumption severs.

Applied to Priya: the clinical guidelines the triage system was trained on were developed in urban tertiary hospitals, primarily in cities, primarily from studies conducted on populations with different diets, different labor patterns, different water access, different disease ecologies. The transfer from those studies to this woman in this village in this drought is not free. It costs something. And the cost is borne by Priya, who receives a recommendation calibrated to someone else’s context.

The non-transferability operation does not say clinical guidelines are useless outside their context of origin. It says the transfer is not neutral. Every portability assumption severs a contextual dependency, and the severed dependency may be the one that matters most for the person in the destination context.

Anti-Categorization

#

The Zhuangzi, one of the foundational texts of Daoist philosophy, is relentless about the limits of categorical thinking. The fish does not know water. The frog in the well does not know the ocean. And you do not know what you do not know, because not-knowing has no signal.

The operation: ask whether the act of categorizing is the right move at all. Sometimes the appropriate response to complexity is not better categories but fewer.

Applied to Priya: maybe what she needs is not a diagnosis. Maybe the entire clinical encounter, the categorization of her life into symptoms and differentials and treatment plans, is the wrong frame. Maybe what she needs is a witness. Someone who sees the totality of what she is carrying and names it without converting it into medical vocabulary.

The system cannot do this. No triage system will ever do this. But the system could flag the moment. The moment when the specification becomes insufficient for the life it is trying to describe. The moment when the correct output is not a better answer but an honest admission: this encounter exceeds my categories. A human who sees differently should be in the room.

I wonder whether the hardest thing about building this architecture is not the technical challenge but the institutional willingness to build a system whose designed output is, in certain cases, the admission that it should not be the one answering.

What the Compound Catches

#

The Intersectional Systemic Harm Index, built from healthcare practice before any of these traditions were consulted, already performs several of these operations without naming them.

It refuses atomization. That is the Ubuntu operation. It treats the interaction between barriers as the real unit, not the individual barrier.

It traces barriers to their compounding consequences. That is the pragmatist operation. It does not care whether each barrier is well-described. It cares whether the compound produces outcomes the decomposed view cannot predict.

It was built from the lived experience of watching systems process people whose lives exceeded the categories. That is situated knowledge made operational. The designers knew the categories were insufficient because they had spent years on the side of the people being categorized.

The traditions give names to what the index does. They explain why it works when conventional assessment does not. Conventional assessment comes from an analytic tradition that decomposes problems into isolable components. The index comes from a relational tradition, arrived at through practice, that treats the compound as the real unit.

Beneath All Seven

#

Roy Bhaskar’s critical realism runs beneath every operation in this essay, whether it is named or not.

Reality, Bhaskar argued, is stratified into three domains. The empirical: what has been observed and recorded. The actual: what has occurred, whether or not anyone observed it. The real: the generative mechanisms that produce events, whether or not those events occur, whether or not anyone observes them.

Each tradition catches a different way the empirical stratum fails to represent the real. Nagarjuna catches reification: the empirical record treats a constructed category as if it were a natural kind. Ubuntu catches atomization: the empirical record treats the individual as isolable when the mechanism is relational. Feminist epistemology catches perspectival bias: the empirical record was produced from a standpoint it treats as universal. The pragmatist catches consequential failure: the empirical record validates the recommendation but not the outcome.

Each is a specific variety of stratum gap. A specific way the surface of the data undershoots the depth of the reality it claims to describe.

The triage system that processed Priya operated entirely at the empirical stratum. It could only work with what had been observed, recorded, published, digitized, and included in its training data. Priya’s life operates at the level of the real, where mechanisms interact in ways no published study has documented because the conditions under which those mechanisms produce observable symptoms in that specific combination, in that specific geography, in that specific household structure, were never the subject of a study.

The gap between the empirical and the real is where the harm lives. The seven operations are seven ways of detecting the gap. None of them closes it. Closing it requires something else: the willingness to work backward from what is actually happening in a life to the mechanisms that produce it, whether or not those mechanisms appear in any existing record.

That is the subject of the essays that follow.

The Nataraja Again

#

Dr. Chandran, the rheumatologist in the previous essay, keeps her brass figurines on the windowsill. Each one is a slightly different rendition of the same pose. The dance of creation and destruction, held in bronze, still and moving at the same time.

Seven traditions. Seven ways of seeing. Each one a different figurine of the same fundamental gesture: the refusal to accept the surface as the whole.

Priya went home with iron supplements she may or may not take. The system recorded the encounter as resolved. The resolution is real at the empirical stratum. At the level of the real, nothing has changed. The water source is still far. The body is still carrying the load. The monsoon is still shifting.

The seven operations would not have changed this. They are not treatments. They are not solutions. They are ways of seeing clearly enough to know that the treatment offered was insufficient for the life it was offered to. That clarity is not nothing. It is the precondition for any intervention that might actually reach the stratum where the mechanisms operate.

Whether the system will be built to see this way, and whether the institutions that deploy it will tolerate being told that their categories are insufficient, is a question this essay cannot answer.

The figurines hold still and move at the same time. The categories are useful and insufficient at the same time. Both are true. The architecture this essay describes is designed to hold both without pretending that one resolves the other.

This is the second essay in The Insufficient, a four-essay sub-series of The Approximate Mind. The first essay, “The Skeptic,” introduced the architectural concept of a system whose resting state is non-belief. This essay populates that architecture with seven philosophical operations, each drawn from a tradition the AI development ecosystem was not built to see: Pyrrhonian suspension, Madhyamaka anti-reification, Ubuntu relational ontology, feminist standpoint theory, pragmatist consequential verification, Indigenous non-transferability, and Daoist anti-categorization. The third essay, “The Intent,” moves upstream from the specification to the commissioning decision and asks who put the categories there and why.

References

#

Madhyamaka Buddhism

Nagarjuna. Mulamadhyamakakarika (Fundamental Verses on the Middle Way). Translated by Jay L. Garfield. Oxford University Press, 1995.

Siderits, Mark, and Shoryu Katsura. Nagarjuna’s Middle Way: Mulamadhyamakakarika. Wisdom Publications, 2013.

Ubuntu and African Philosophy

Ramose, Mogobe B. African Philosophy Through Ubuntu. Mond Books, 1999.

Metz, Thaddeus. “Ubuntu as a Moral Theory and Human Rights in South Africa.” African Human Rights Law Journal, vol. 11, no. 2, 2011, pp. 532-559.

Feminist Epistemology

Harding, Sandra. Whose Science? Whose Knowledge? Thinking from Women’s Lives. Cornell University Press, 1991.

Collins, Patricia Hill. Black Feminist Thought: Knowledge, Consciousness, and the Politics of Empowerment. Routledge, 2000.

Hartsock, Nancy C.M. “The Feminist Standpoint: Developing the Ground for a Specifically Feminist Historical Materialism.” In Discovering Reality, edited by Sandra Harding and Merrill B. Hintikka. Reidel, 1983.

American Pragmatism

Peirce, Charles Sanders. “The Fixation of Belief.” Popular Science Monthly, vol. 12, 1877, pp. 1-15.

Dewey, John. Logic: The Theory of Inquiry. Henry Holt, 1938.

James, William. Pragmatism: A New Name for Some Old Ways of Thinking. Longmans, Green, 1907.

Indigenous Epistemologies

Smith, Linda Tuhiwai. Decolonizing Methodologies: Research and Indigenous Peoples. Zed Books, 1999.

Kimmerer, Robin Wall. Braiding Sweetgrass: Indigenous Wisdom, Scientific Knowledge, and the Teachings of Plants. Milkweed Editions, 2013.

Daoist Philosophy

Zhuangzi. The Complete Works of Zhuangzi. Translated by Burton Watson. Columbia University Press, 2013.

Critical Realism

Bhaskar, Roy. A Realist Theory of Science. Verso, 1975.

Bhaskar, Roy. The Possibility of Naturalism: A Philosophical Critique of the Contemporary Human Sciences. Harvester Press, 1979.

Category Theory and Classification

Bowker, Geoffrey C., and Susan Leigh Star. Sorting Things Out: Classification and Its Consequences. MIT Press, 1999.