Wednesday, 15 July 2026

RESEARCH — Using AI to See What Perception Presupposes


How we see shapes what we see

Each field polishes its own bead.
This research asks where those beads begin
— a different direction.

 

Background & Method

As AI use becomes universal, something once invisible is beginning to come into view: the preconditions of perception. Like air, which surrounds us without being noticed, there is a matrix that shapes how we see while remaining unseen — and that matrix is the precondition of perception.

Different AI models, for instance, give different answers to the very same prompt. Even within one model, a shift in the reasoning version alone changes the answer again. This difference is not a matter of tone or style. It is because, before the question is even asked, the matrix that produces the answer is already woven differently. AI is not the only case. A matrix of this kind underlies our perception as well. This matrix is like a lens that bends what we perceive. We cannot look at this lens, because we look through it.

The same holds for scholarly research. Modern research, growing ever more precise and solid within a single field, can be compared to bead-making. But to see the world as it is, three things are needed together: making and refining beads, stringing beads, and knowing the membrane — the matrix — that forms each bead. If reductive research is bead-making, and interdisciplinary research is bead-stringing, then this research is the work of grasping the membrane that constitutes each bead.

This membrane is not a wall enclosing the bead; it refers to the way of seeing itself. We do not look at the membrane; we look through it. And this research is no exception. The work of grasping the membrane is itself carried out from within another membrane. Only its direction differs — it faces not the detail inside the bead, but the point where beads take form.

To an eye that sees through the membrane, the membrane itself is as invisible as air. Even without seeing it directly, sensing its presence requires first withdrawing attention from the details inside it — and that withdrawal inevitably means missing those details. So to a specialist in each field, this work may look coarse and inaccurate, and that criticism is largely fair.

Still, this work is necessary. Bead-making alone cannot reveal how scattered beads connect, so seeing the whole calls for stringing them. And knowing the conditions under which the membrane forms is the groundwork that lets us properly see the points where beads meet when we string them.


Roadmap

This research approaches the invisible preconditions of perception from three directions: 

1/ WHAT is invisible?

2/ HOW does it operate?

3/ WHY is it invisible?


1/ WHAT is invisible?

When different AI models give different answers to the very same question, where does that difference come from? Here, the research traces the divergence back to its boundary and looks for signs of a precondition already at work before the answer was produced. In doing so, it treats AI — built on human cognitive research — as a partial mirror that reflects certain structures of cognition. Just as biology can study aspects of human nerve function through C. elegans even though humans are not worms, there are aspects of human cognition that can be studied through AI as a structural model, even though AI and human cognition are not identical.

This invisible operation goes by different names depending on which aspect is in view. When its invisibility is the focus, it is called a precondition. When the emphasis falls on the space in which the unseen mechanism operates, it is called a matrix. When what matters is the point at which it divides one dimension from another, it is called a boundary or a membrane. This shifting of names is not carelessness but the very perspective of this research — names change depending on which facet of a single operation is being captured, and the lens of each vantage point shapes what is perceived.

Papers (about precondition)

  • [Published] Buddhism and Agency in the Age of AI (2026) 불교평론
  • [Restructuring] Diverging Perceptions: Why We See Differently
  • [Restructuring] We See the Model, Not the World — What AI Ontology Reveals About the Conditions of Perception
  • [Planned] Self as a Boundary


2/ HOW does it operate? 

Knowing that an invisible precondition exists is not the same as knowing how it works. Here, the research asks what structure this precondition takes, what effects it produces, and what it would take for its operation to cease so that an awareness of the precondition as precondition can emerge. It pursues these threads with information theory as its model. In this way, the work arrives at the very end-point of the cognitive operation that gives rise to the precondition of perception, and sets out to dismantle the subjectivity that sits there.

Papers (about mechanism)

  • [Under Review] When Speed Meets Speed: AI-Induced Cognitive Overload
  • [Restructuring] From Rigidity to Collapse: A Triadic Structural Analysis of Information Theory
  • [Restructuring] Pattern Analysis of Negation Across Traditions
  • [Planned] From Loop to Exit: The Triad of Time in 0 and 1


3/ WHY is it invisible?

This precondition is not invisible simply because it is transparent. The very way language categorizes the world makes it structurally unable to disclose this precondition. Here, the research addresses the inevitable distortions that arise when a spectrum is categorized and represented, and re-reads techniques of negation such as avyākata, neti neti, and apophasis — not as mystical residues, but as methods for actively undoing categories that have hardened into the wrong shape.

Papers (about categorization & representation)


In A, the question is whether this precondition — never before recognized as a precondition — exists at all, and if so, in what form it reveals itself. In B, the question is how it operates. And in C, now that its form and operation have come to light, the question is why it could not be perceived until now.


The papers on this roadmap are divided and restructured from an unpublished manuscript, Reframing Existence: Tracing Patterns Beneath Language.


May this reach someone who sees the same direction


at