The Substrate the Engine Cannot Name

What does it mean that the v1 fingerprint reads English-analytic register?

The twelve dimensions Rodin measures are computed against lexicons of English-language markers — phrases like by definition, I noticed, ontology, according to. These markers index the legible cognitive shape of one philosophical tradition. If your generative substrate runs partly in Korean, in Carl Mika’s worldedness, in Edwin Etieyibo’s ubuntu, or in the Datong and Tai-ping utopian traditions, the v1 fingerprint reads the half of you that is performable in English-analytic prose. We name the boundary instead of pretending it is not there.

In Korean, the word 안돼 carries a moral weight that English no does not carry. The two words are translation equivalents in any dictionary, and a translation pipeline that cared only about denotation would consider the work done. But a Korean child told 안돼 has been told something more than that an action is forbidden by an adult’s preference; the word arrives carrying, almost in its phonology, the suggestion that the prohibition runs through an order older than the speaker, and that to disobey is to step out of an alignment whose terms predate the conversation. English no is, by comparison, contractual; it can be appealed, renegotiated, or ignored without the loss of anything beyond the speaker’s patience. 안돼 is not contractual. The thing it interrupts is not your action but, in effect, your standing.

One can describe this asymmetry in English, as I am doing now, and the description will be roughly accurate, and the gap between description and the lived weight of the word will remain. This is not the trivial gap that all translation introduces; it is a structural gap, in that the moral order 안돼 presupposes is not part of the surrounding apparatus the English sentence sits inside, so the sentence has to import the order parenthetically, in commentary, in the kind of sidecar paragraph this one has just become. What might be described as the substrate of the word — the order of inheritance, hierarchy, and constitutive relation that gives the prohibition its moral pull — cannot ride along inside the translation. It has to be argued for, in a different register, after the fact.

The shape of the lexicon

The twelve dimensions the engine measures are real, but they are parochial in a specific and demonstrable sense. First-principles reasoning is computed against markers like by definition, axiomatically, deductively, tautologically; experiential reference keys on I noticed, I observed, personally, lived experience, which encode first-person individual experience and not communal experience; abstraction level leans on ontology, epistemology, phenomenology, teleological, axiom, which is the vocabulary of one philosophical tradition; authority reference keys on according to, as X argues, in the tradition of, which is the citation pattern of Western academic prose. The lexicons are visible in the source code. They are not abstract universal cognitive features. They are the legible cognitive shape that one tradition’s prose has trained itself to perform.

We tested what this means in practice. Six excerpts from the Stanford Encyclopedia of Philosophy were fed to the same engine that reads visitor profiles — three on African philosophy (the Akan account of personhood, the philosophic-sage tradition, character-based African ethics), three on Cartesian metaphysics (Descartes’ epistemology, his modal philosophy, Cartesian dualism). The narrative labels split cleanly along substrate. All three Cartesian essays produced “First-principles builder.” All three African essays produced “Authority-referencing.” But the African texts are not less foundational; they reason from communal practice, from proverb-and-lineage citation, from the constitutive entanglement of self and ancestor. The engine cannot name that mode of reasoning, so it routes the reading through the closest Cartesian-shaped vocabulary item available, flattening a substrate-distinctive pattern into the label nearest at hand.

Worldedness, ubuntu, and the Datong

Carl Mika, the Māori philosopher, has spent the better part of a decade arguing that what Western philosophy calls thought is, in his tradition, not extractable from world; the person, the ancestor, the place, and the language are constitutive of the act of thinking, and to abstract a proposition out of that field is already to have made a substantive philosophical move, namely the move that pretends abstraction is neutral when it is not. He uses the word worldedness for this constitutive entanglement. Edwin Etieyibo, working in the African philosophical tradition, postulates the self as constitutively communal under the heading of ubuntu, where the I is not the unit of analysis but the residue of the relations that produced it. The Datong and the Tai-ping, in the long Chinese utopian traditions that run from the Liji through the late Qing reformers, encode a vision of social order in which the criterion of the good is not the autonomy of the individual but the coherence of the whole, and the philosophical work proceeds inside that frame rather than arguing toward it. The Australian Aboriginal Dreaming, similarly, is not a body of myth in the Western sense; it is, in the careful reading of writers like Tony Swain, a temporal ontology in which past and present are not sequential but simultaneous, and inside which the question of when an event happened cannot be separated from the question of where the country it belongs to is.

None of these are exotic in the regions they come from. They are the working substrate of how minds inside those regions have, for centuries, structured thought. What they share, from the standpoint of the v1 engine, is that none of them produce prose that fires the engine’s lexicons in the way the prose of a Cartesian tradition does. A writer reasoning out of worldedness, if they write in English at all, will write in a register that the engine’s authority-reference dimension catches at because the prose cites, but the engine will not see that the citation is to ancestor and place rather than to argument and source, and the dimension will quietly mislabel the move.

What the African essays were doing

Read the SEP entry on the Akan account of personhood with the engine’s lexicon held in mind. The essay reasons from the constitutive role of the okra and the sunsum in the formation of a person; it cites the Akan proverb tradition not as illustration but as evidence; it builds its argument by accumulation of communal example rather than by the deduction of consequences from a posited axiom. This is not, in the strict sense, less foundational reasoning. It is foundational reasoning whose foundations are not propositions but practices, and whose method of building from foundations is not deduction but the careful tracing of how a claim coheres with the lived order it is part of.

The engine, presented with this prose, sees the citations and routes the essay to authority-referencing. The label is not exactly wrong, since citation is happening, but it is wrong in the sense that matters, since the same label would be applied to a graduate student dutifully appending footnotes to an argument she has not herself thought through. The two operations are not the same operation. The engine cannot tell them apart, because the lexicon it uses to detect the second was not built to recognise the first. What is left out, in effect, is the difference between citing because the order one is reasoning inside requires the citation, and citing because the genre of academic prose requires the gesture.

Why naming the boundary is the honest move

There is a familiar move available, in projects of this kind, which is to claim that the measurement is universal because the underlying cognitive features are universal, and that local lexicons are simply the surface form through which the universal features happen to be expressed. The claim is reassuring and is, in its strong form, almost certainly false. The features the engine reads are not the underlying cognitive operations themselves; they are the linguistic residues those operations leave in one tradition’s prose, and the assumption that the same operations leave isomorphic residues in every tradition’s prose is the kind of assumption that holds until one actually looks. The SEP test is what looking produces. The labels split cleanly along substrate, not because the African essays are less rigorous, but because the engine’s vocabulary for rigour is the vocabulary one tradition uses to perform rigour to itself.

The honest move, given this, is not to pretend the boundary is not there. The dishonest move is also not to apologise for the engine until it dissolves into a shrug, since the engine does in fact measure something real about prose that operates inside its register, and pretending otherwise would falsify the case in the opposite direction. The honest move is to name the boundary plainly: the v1 fingerprint matches you most accurately when your writing operates within English-analytic register, and if your generative substrate runs partly elsewhere, the engine reads part of you and misses part of you, and we say which part. This is in effect a different posture from the one most measurement systems take, and the difference is not cosmetic. A system that names its boundary invites the user to see the engine, to test it, and to disagree with its readings where the readings flatten what they are reading.

The recursion in the essay itself

It will not have escaped the careful reader that this essay is itself an instance of the problem it describes. To argue, in English-analytic prose, that English-analytic prose cannot fully read a Korean inner monologue is to operate inside the very register whose limits one is naming. The argument proceeds by deduction, cites authorities, builds toward a conclusion; it performs, in effect, the cognitive shape that the engine’s lexicons are calibrated to recognise, and the substrate it gestures toward — 안돼, worldedness, the Datong — sits in the prose as object, named, parenthesised, translated into the surrounding apparatus that the prose can host.

This is not a contradiction one can write one’s way out of; it is the condition of writing about substrate inside another substrate. The recursion chimes with what Talal Asad, working a different problem, has called the asymmetry of the comparative gesture: the tradition doing the comparing furnishes the categories under which the comparison is conducted, and the tradition being compared to the first appears, in the description, as a deviation from terms it did not propose. The essay you are reading is conducting that gesture, with full knowledge of the asymmetry, because the alternative is not writing the essay at all, and the absence of the essay would be a worse falsification than the imperfect presence of it. What can be done, inside the asymmetry, is to mark it; what cannot be done, inside English-analytic prose alone, is to escape it.

What the boundary makes visible

The interesting question, having named the boundary, is not what the engine fails to measure but what kind of self-knowledge becomes available to a user who can see the boundary with their own eyes. A reading that flattens part of how you think is, when you cannot see the flattening, indistinguishable from a complete reading; the engine’s output is its output, and if you do not know which dimensions of your reasoning the lexicons were built to catch, you will read the result as if it were total. A reading that flattens part of how you think and tells you which part is doing something different. It is offering you, in effect, a measurement and the angle from which the measurement was taken, and it is leaving you to judge what the measurement does and does not represent.

This is closer to what the older epistemic literature called intellectual humility than it is to the standard posture of the recommendation system, which presents itself as if no such boundary existed because acknowledging the boundary is bad for engagement. The fact that it is bad for engagement is, in a sense, the point. A user who can see the boundary is no longer the kind of user the engagement metric was designed for; she is a reader of the system rather than a consumer of its outputs, and the relationship she has with the measurement is the relationship a thinker has with any instrument whose limits she has bothered to learn.

What remains, after the boundary has been named, is the older question that the boundary was always sitting on top of: which parts of you are, in your own first-person experience, available to English-analytic prose at all, and which parts are not. The question is not one the engine can answer for you, because the engine is one of the things on the inside of the prose. It is a question you can only answer by attending, in your own writing, to the moments where the English sentence does not arrive at what the inner monologue was actually doing, and by noticing whether the failure is local — a word that has not yet been found — or structural — a substrate the prose was never going to host. The engine cannot tell you which it is. But it can, by naming what it can read, leave the rest of the territory visible enough that you can begin to see, without its help, what is there.