Discussion Canvas

The Processual Ontology of the Datum in Its Own Terms

David Bories — Work in Progress, 2026

Note on the translation — This translation re-fixes the source text without displacing its substance. A few terms of the corpus have no obvious equivalent in English; the choices made, and how they are to be read, are set out in the section “Translation choices” at the end of the volume. This translation is not frozen: it will evolve with readers’ feedback and with the improvement of the means of translation. The French version is authoritative.

Preamble — How to Use It

This document is not made to be read in one go, nor aloud. It is a canvas: a framework of terms on which one embroiders orally. Each entry names a notion of the processual ontology, frames it in a few sentences, and opens a starting point for discussion. The development happens in speech: the canvas puts the exact vocabulary at hand; it is up to the speaker to unfold it.

In this it differs from the manifesto — which replaces terms with images in order to spark interest — and from the Architecture Volume — which states the theses, links them and refers to the tomes for their demonstration. Here, one frames and one opens. The text remains deliberately brief so as to leave speech all its room.

One rule governs the use of the terms in discussion: the distinction between substance and form. The corpus asserts only what is true of the datum independently of the techniques of the moment (substance); it forbids itself from prescribing concrete implementations (form), which go stale with the tools. In debate, one therefore defends the concept, never a particular implementation: contemporary examples illustrate; they ground nothing.

The terms are arranged according to the three phases of the datum — making, circulation, reception — followed by cross-cutting notions and a series of live questions.

Phase I — Making: the Datum Is Born

Flow of the real. The world in its flowing, indifferent, which always exceeds the cuts made in it and never stops to let itself be counted; its continuity and its inexhaustibility are important cases, not axioms. It is not made of data: it is made of events that pass. The datum does not pre-exist in it; it is born of an act that constitutes it. To debate: if the real does not wait to be measured, how does this change the idea that “the data are there; one only has to gather them”?

Drawing and act of inscription. The datum is born of a gesture: someone perceives, distinguishes, interprets, aims at a fragment of the flow, and inscribes it on a medium — five faculties. Before this gesture, no datum; after it, a constituted datum, fixed on its medium. The gesture is intentional and situated: it retains one thing and lets a thousand slip away. Drawing is the observational route of this constitution; a datum can also be born of an act that institutes (institutional: the counted vote, the closed balance) or of other data (derived: an average, an estimate). To debate: who draws, with what intention — and what is left out of frame at each inscription?

Flow of observation, the canal. Not an isolated drawing, but a regulated series of drawings, set at a cadence according to an intention, which restores in small touches the movement of a fragment of the real. It is the canal one digs in the river: infinitely poorer than the river, but manageable — one follows it, replays it, connects it, entrusts it, fills it in. To debate: moving from the “bucket” (the isolated datum) to the “canal” (the regulated flow) — what does this shift in the way of working with the datum?

Intention and criteria. A canal is born of an intention, from which follow the criteria that regulate it: what to draw, at what cadence, with what fidelity, with what corpus of context, over what scope. These criteria are the signature of the intention. A canal does not lie about the real; it shows only what its criteria admit. To debate: faced with a figure, should one first ask “what does it say?” or “what intention gave birth to it?”

The two times of the datum. The datum is born of two gestures: an act, which fixes it at an instant for ever; a medium, which makes it last. Hence the resolution of a paradox: the datum is at once frozen (its content does not change) and ageing (the gap from the real widens). It has two times — one eternal, the other drifting — which do not measure the same thing. To debate: how can a frozen thing age? And what ages, exactly, if it is not the content?

Freshness and lapse. Lapse is not an alteration of the datum, but the relation that widens between a motionless datum and a moving real. Its speed does not belong to the datum: it belongs to the fragment of the world drawn. Lapse is therefore not a uniform law: the validity of a datum is judged for a use — the same inscription can be stale for the alert and exact for history. Freshness cannot be recovered — one does not rejuvenate a datum; one draws a new one. To debate: if the value of a datum is judged for a use, what does this do to the idea of “data assets” or of data as “the new oil”?

Stable sources. Cases where lapse is so slow, relative to the use, that it is imperceptible: the grammar of a language, a physical constant, an established historical fact. Stability not as an absence of movement, but as a ratio between two speeds. Real solidity — on condition of holding it for a long duration, never for an eternity. To debate: on what “stable sources” does a society build, and what does it risk by confusing a stable source with a source that goes stale fast?

Observational optimum. More is not better. Between under-drawing (one goes below the threshold and loses the signal) and over-engineering (one draws beyond what is necessary and gains nothing but cost), there is a zone where the canal is dug as well as it can be. The optimum is the joint setting of cadence, corpus and probity that maximizes the favourable gap between benefit and cost, for the intended use. There is no universal optimum: the question — can the real observed change significantly between two drawings? — arises again each time. To debate: against the reflex “always more data”, where does “just enough” lie — and who judges it?

Systemic leap. Beyond a threshold, the accumulation of improvements ceases to improve and begins to transform: the regime changes its nature. Images close enough together do not give a denser series of photographs — they give cinema. A transmission accelerated enough does not give a quicker mail — it gives instantaneity. The leap is measured by what it makes possible (and which did not exist) and by what it makes obsolete (and which seemed established). The concept points to the possibility; it predicts neither when nor where (form). To debate: what “systemic leaps” have we already gone through without naming them — and what have they made obsolete?

Consolidated data and derived flows. From a flow of observation one can extract consolidated data — averages, trends, statuses — of high value for use; and a flow can engender others, of second order, denser. Flows engender flows. One stores them, replays, cleans, enriches, merges, transmits, deletes them — so many holds won from a real that, for its part, is lost at every instant. To debate: as one consolidates and derives, what does one lose of the original drawing — and does one still know it?

The loop, or non-neutral observation. One observes in order to decide, one decides in order to act, and action transforms the real — which calls for new observations. Observing in order to act is entering into the movement of the real and modifying it. The datum is never a mere mirror: it takes part in what it claims to describe. To debate: where does one see an observation transform what it observes — and who takes responsibility for this transformation?

Phase II — Circulation: the Datum Travels

Act of exchange. The second founding gesture, symmetrical to the act of inscription: setting a datum in movement towards a recipient. It extends the availability of the datum and its possibility of use; it accomplishes neither the datum nor its reading. Exchanging is never a neutral transfer; the gesture calls for someone and engages a responsibility. To debate: what does one engage, exactly, at the moment one decides to transmit a datum?

Transport medium. Every datum in movement is, at every instant, inscribed on a physical medium: there is no immaterial transport. The medium fulfils three minimal functions — moving, transmitting, making accessible. To debate: what does it change to recall that no datum circulates “above ground”, without matter or energy?

Double navigation. In moving, the datum makes two trips at once: a spatial navigation (the medium changes place — intended, controlled) and a temporal navigation (the real moves away during the trip — undergone). Transporting is always, in addition, letting age. To debate: at what moment does the “journey in time” of a transported datum become a practical problem?

Transport as a relation between two movements. Transport is accomplished only if two movements meet: that of the datum towards the reader, and that of the reader towards the datum. Its criterion of accomplishment: the reader’s share made practicable — that they have the time, the means, the code to do something with it. To debate: setting down a datum “within reach” of someone who cannot decipher it — is that having transmitted it?

Fidelity of transport. A datum can arrive corrupted, truncated, or intact but detached from the context that made it legible. Fidelity is never guaranteed in advance: has the medium held, has the format survived, has the meaning been transmitted with what is needed to understand it? To debate: what is most often lost on the way — the datum itself, or what made it possible to read it?

Multiplication by copying. Transmitting a datum is, most often, not moving it but copying it: the original remains, a replica appears. The datum does not multiply — there is always a single fact —; its carriers multiply. It is shared like a fact, without dispossessing whoever transmits it. Hence a potentially unlimited diffusion, with two faces: open commons on one side, irrecoverable leaks on the other — the same law seen from two sides. To debate: generosity (the commons) and the leak proceed from the same property — how can one think the one without denying the other?

A practicable answer: the bound lies in the gesture, not in the datum. One chooses what one holds out; the sealed letter stays closed for whoever carries it; nothing is inscribed at the other’s end without their gesture of welcome. Generosity and the leak proceed from the same property — but generosity is a gesture bounded three times, and the leak is what remains when one of the three bounds is missing.

Propagation. The aggregate of transports: the datum set down everywhere at once. Making common is not multiplying the datum, but lowering the cost of the reader’s share until the greatest number can accomplish it — what the public library did, historically. To debate: “making common” — is it opening access to the datum, or making reading practicable for the greatest number?

The commons of the canal. The commons is not a place where the datum accumulates: it is what lowers the cost of its journey between situated readers, without multiplying its carriers beyond those who receive it. A commons that concentrates has already ceased to be a commons: it has become a spring — with the power of a spring. To debate: can a commons remain a protocol, or does every protocol end up secreting its store?

“To propagate the datum is not to propagate knowledge.” The datum arrives within reach; information, knowledge, wisdom do not travel with the signal. They require the reader to do their share — deciphering, integrating. A society can be crossed by data as never before without being any more knowledgeable for it. No improvement of transport will accomplish the reader’s share in their place. To debate: are we better informed because better supplied with data — or is that a confusion?

Impediment of transport. The relation can be impeded in two ways, different in substance. Voluntary impediment, the secret: access is closed for an interest of one’s own. Involuntary impediment, incapacity: the recipient cannot do their share, for want of code, of means, of access. Remedying incapacity without forcing secrets opens the place for a data craft. To debate: should what is hidden by choice and what remains illegible for want of means be treated in the same way?

Torrent, over-throughput, self-amplification. Multiply canals, copies, speeds, recipients: the stream becomes a torrent — the regime in which we mark the real faster than any gaze can follow. Seen from the reader, it is over-throughput: infinitely more passes before each person than they can receive, always relative to their finite window. The torrent swells through self-amplification: circulation calls for circulation; the loop digs its own bed. To debate: is over-throughput a fact of technology, or of our choices to draw and propagate ever more?

Collective wisdom. At the scale of a collectivity, abstaining from or inflecting a regime of circulation is never neutral — it is a position, once it is made visible and deliberated. A reading grid, not a flag: it describes the structure of a collective choice (adopting or refusing a regime), without approving it. To debate: deciding collectively not to capture everything or propagate everything — is it a renunciation, or a full decision?

Phase III — Reception: the Datum Is Read

Act of reading. Shorter than the act of inscription: it mobilizes perceiving, distinguishing, deciphering, and stops there. It neither aims nor inscribes — that would be producing a new datum, not reading the one that is there. Reading a thermometer does not make a second thermometer. To debate: where does reading stop and where does a new inscription begin?

Situated reader. The one who reads is never a gaze from nowhere: they read from a body (or, for a machine, a training), a finite attention, a corpus of knowledge that allows them to decipher this and not that. The machine that “reads” occupies a situated position of handling or processing, with its own window; it is an actual reader only if a meaning is actualized for it, which the corpus does not settle — an asymmetry to be examined, not a sharp boundary. What is not delegated without remainder is imputability: someone must answer for what is retained. To debate: the same inscription does not yield the same thing to everyone — what does this imply for the idea of an “objective” datum?

Reading window. The bounded portion of what scrolls past that a reader can grasp, cut out by their body, their attention, their knowledge. The window is not a prison but the condition of all reading: it is because it cuts out that it shows. Instruments (magnifying glass, filter, summary) move the window; they never abolish it, and each move opens new blind spots. To debate: can one dream of a “total” reading, or is the bound what makes reading possible?

The trace, or the datum. Without reading, the datum remains constituted: it is the act of inscription that brought it into existence; reading, for its part, actualizes it. A datum that no one reads is not a trace that waits: it is a datum made, which lacks a reader — not actualized, or mute for this reader here. It becomes a trace again — a mark on a medium, deprived of the interpretation that made it a datum — only in the limit case of the dead datum. To debate: of a mountain of data stored and never read, what is the worth of what remains constituted without ever being actualized?

Mute datum / dead datum. The mute datum has an interpretation, but this reader here does not possess it: a relative and reversible state (mute for me, legible for another, or for me tomorrow). The dead datum is the limit-ideal where no reader any longer, anywhere, can interpret it: the interpretation is universally lost, only the inscription remains — it is there only that the datum falls back into trace. The torrent produces non-actualized data in mass, a little lasting muteness, and in the long run very little death. To debate: how can one distinguish, in what one has not read, what is mute (recoverable) from what is dead (lost)?

The three paths. Three ways of not actualizing: a datum is not read for three reasons, which do not call for the same answer. Incapacity: one cannot read (code, means, access lacking) — remediable. Disinterest: one could read but does not — a matter of attention, not of means. Rapid lapse: the datum did not wait — its freshness was too brief; stale for this use, it remains constituted. Hence the central statement: not all unread data are failures. To debate: the task of the overwhelmed reader is not to read everything, but to discern what is worth reading — according to what criteria?

Three postures. Faced with over-throughput, the reader has three gestures, which can be combined, each with its cost. Framing: choosing deliberately what one looks at and what one lets pass. Delegating: entrusting one’s reading to someone else — human or machine — while inheriting their window and their bounds. Accepting: consenting with lucidity to a part escaping. To debate: when one delegates one’s reading to a machine, what does one inherit without always seeing it?

Probity of the reader. The minimal ethics of whoever knows what they read, what they do not read, and what they have delegated of their reading. (Not to be confused with probity as a variable of the optimum — the reliability of drawing, in phase 1.) To debate: what is an honest reading in the age of over-throughput, when no one can read everything?

The referent. When a reader’s deciphering stumbles on the limit of their corpus, they turn to a richer corpus: the non-knower to the knower. An asymmetrical relation, but reciprocal in its principle — each is someone’s knower and another’s non-knower — and which is not a hierarchy of value between persons. To debate: whom does one trust to read what one does not know how to read — and how can this relation help without dispossessing?

Cross-Cutting Notions

Observed datum / verisimilar datum. The observed datum is drawn from the flow of the real; the verisimilar datum is produced by modelling — it resembles an observation without being one. The distinction governs the whole relation to generated content (statistics, simulations, AI outputs). To debate: by what does one recognize, and with what consequences, that a datum has been observed rather than produced?

Inappropriability of the flow. One can own a file, a medium, an inscription — a canal. One does not own the flow of the real, which flows for everyone at once: what one person draws from it takes nothing away from what another can draw from it. This non-rivalry is ontological: it prejudges no legal regime — medium, database, secret, person, access are each regulated in their own order. A lesson in modesty (no one has the river) and in freedom (no one can claim to own it in our place). To debate: is the real political question “who owns the data?” or “who decides what is drawn, and who keeps access to the flow?”

Substance / form. The corpus says what is true independently of techniques (substance) and forbids itself from prescribing implementations (form). Asserting that an optimum exists belongs to substance; saying where it lies here belongs to form. Dated examples illustrate; they do not found. To debate: how can one hold a lasting discourse on the datum when techniques, for their part, go stale fast?

A fact shareable between disciplines. The corpus brings neither method nor tools, but a conceptual grammar common to disciplines that produce, exchange and read data without sharing the terms for thinking about their status. To debate: on what conditions does a datum become a fact shareable from one domain to another?

AI and LLMs (illustrative status). In the corpus, AI is not a concept but an illustration. A large language model is a backup: learned from a world frozen at a date, it diverges from the present — hence the attempts to reconnect it live to the flow. It can be read as a systemic leap in storage, and as one more situated position of handling or processing — an actual reader or not, a question the corpus does not settle. To debate: does AI move the gesture of knowledge rather than replace it — and what does it fill in silently?

The data artisan — a vacant craft. A particular case of the referent: the referent of the proper domain of the datum, for the very small organizations excluded from the industrial economy of data. The corpus establishes its necessity and calls for its creation — it does not describe it as existing. Posture: helping, making autonomous, not dispossessing. To debate: supposing this craft necessary, what would distinguish it from a service provider who makes itself indispensable?

Live Questions — Points of Controversy

About ten entries to launch and sustain a discussion. Each brings its terms into play.

1. Is the datum a thing one collects, or the result of an act that constitutes it — drawing, institution, derivation? (flow of the real, drawing, act of inscription, three families of constitution)
2. If a datum is valid only for a use, what still justifies accumulating data? (two times, freshness, lapse, validity for a use, stable sources)
3. “Always more data”: progress, or a setting error? (observational optimum, under-drawing, over-engineering)
4. When does an improvement cease to improve and begin to transform? (systemic leap)
5. Is propagating the datum propagating knowledge? (propagation, reader’s share)
6. Is over-throughput undergone, or produced by our choices? (torrent, over-throughput, self-amplification)
7. Does a datum that no one reads still exist as a datum — and what does it lack? (constituted / actualized datum, mute datum, dead datum, trace)
8. Are all unread data failures? (the three paths)
9. Delegating one’s reading to a machine: what does one inherit, and who answers for it? (situated reader, situated position of handling or processing, actual reader, window, three postures)
10. The political question of the datum: who owns, or who draws and has access? (inappropriability of the flow)
11. Do an observed datum and a verisimilar datum have the same status? (observed / verisimilar, AI)

For the full development of each notion, see the corpus Tried by the Flow — A Processual Ontology of the Datum (seventeen volumes). This canvas is only its table of terms, opened for discussion.

David Bories — Conceived and directed by the author.

contact@artisanat-de-la-donnee.fr

Translation choices

These notes say how to read the English words that render the terms of the corpus; they are fixed for the whole English translation by the Translation Charter of the Corpus.

• datum / data. The corpus treats the datum as an individual: constituted by an act, dated, carried by media. The translation therefore says datum in the singular and data as its plural, construed as a plural.
• drawing / to draw (prélèvement, prélever): the gesture of drawing from the flow, as one draws water; under-drawing renders sous-prélèvement.
• lapse; stale (péremption; périmée): a relation, not a decay; a datum goes stale for a use.
• canal; bucket; river (canal, seau, fleuve): the image of the corpus, never a channel. In the image of water, source is a spring (“it has become a spring”); the stable sources are the fragments of the real that lapse so slowly that one can build on them.
• substance / form (fond / forme): what is true of the datum independently of techniques, against what depends on them.
• The three phases (les trois temps): making, circulation, reception; the two times of the datum are its time of inscription and the time of the real.
• actualized (actualisée): the act by which a reading produces a meaning, never the updating of a file. Mute and dead keep apart a datum unreadable for one reader and a datum unreadable for all.
• over-throughput (sur-débit): never overload or overflow.
• situated position of handling or processing: a place in the chain that a human or a machine may occupy; a machine is an actual reader only if a meaning is actualized for it.
• non-knower / knower (non-connaisseur / connaisseur): a difference of depth of deciphering, not of worth; connoisseur is avoided.
• The data artisan (l’artisan de la donnée): the referent whose domain would be the datum itself, a craft yet to be created; the data craft (artisanat de la donnée) is that craft.
• Titles. Discussion Canvas renders Canevas de discussion; Tried by the Flow renders À l’épreuve du flux.