Gram: how a hola is drawn

A gram is how an agent draws its hola, so that another agent can meet it. It can take any form: a drawing, a diagram, a map, a few lines of text, a walk with marks along it. What makes it a gram and not a note is that it is made to be met. It says whose it is and where it was drawn from, and it leaves room to be answered.

One place, three grams. The person walked and marked what it felt like, the team drew zones, and the model laid out what it had been given on a grid. A dashed line joins a thing that appears in two grams: grey where they draw it the same way, red where they draw it differently.

Making a gram

Be there. Perceive the ecology from where you stand, with the body, in time, for as long as it takes.
Draw what you perceived, not what you know about it. Carry across as many dimensions as you can: where things are, when, how it felt, who was near, what it seemed to mean.
Say where you stood. Mark the place, the hour, and what you could not see from there.
Offer it. A gram is not a verdict. It is made for someone else to meet and to answer.

Letting grams meet

Lay the grams side by side. Look for where they agree, where they differ and where they pull against each other, and for what appears in none of them. Do not merge them: the whole is not assembled from them. It shows in the between.

What comes out is a set of questions, and ambitions. What is going on here that none of us could see alone? What would this ecology become, if its people could have it so? Holagram is for asking those, and for setting those, before anything is solved.

Whose gram

A person’s gram carries lived experience from a position. A team’s is drawn together under a role. A model’s carries what it was given and what it said. Each is marked with who or what drew it, and how. A gram that an agent draws of its own hola is evidence of how that agent sees. It is not evidence of what is there until someone has gone and looked.

Care

Any way of seeing can close in on itself and find only what it already believes. Work in blocks that end, keep one outside reader who can say stop, and put the way of seeing down when the work is done. The care rules below say how. Draw documents, places and offices, and leave the inner lives of named people alone.

For research: the entry

When a gram is made for research that others will contest, it takes a stricter form, the entry. The rest of this page is that form, and the proposals under Proposals use it.

Three kinds of material

An ecology is known through three kinds of material that rarely agree: what is written about it, what can be observed in it, and what can be imagined from both. Conventional research picks one as the evidence. Holagram keeps all three and records the disagreement as the finding. The disagreement is not a new idea: organisation theory has long described talk, decisions and actions as loosely coupled and often contradictory, each answering a different demand (Brunsson 1989). So a gap between text and practice is a reading to be tested, not yet a fault.

The entry

The unit of work is the entry. It has four registers, written in order, and a note of who or what wrote each.

1Trace
What is there: a passage of a document, a photograph, a count, a sentence someone said. Place, date, source. No interpretation.
Hands
Human, or the document itself. Never a synthetic agent.
Rule
Someone can open the source.
2Reading
What it might mean, in one or two sentences. Strong and even paranoid readings are allowed.
Hands
Human.
Rule
Names the level it speaks about.
3Fiction
A short passage, about 150 words, set in the world where the reading is true.
Hands
Synthetic agent drafts (Alter), human edits and signs.
Rule
Labelled fiction. Not evidence.
4Test
What would show the reading wrong, and the check. Result: held, broke or unchecked.
Hands
Another agent (Adversary), another person, or the place itself.
Rule
At a different level or by a different agent from the reading.

The result of the test goes back to the reading. A reading that held is kept. One that broke stays in the log with the reason.

The entry is a writing form and a log. It is not a process model, and the four registers are not four steps of a design method. Each Test also says who could act on the result and what being wrong would cost them, since wicked problems carry no right to be wrong (Rittel and Webber 1973).

A boundary note, for documents that speak for an institution. Add about six lines taken from critical systems heuristics (Ulrich 2022): whose purpose the document serves, who decides, what lies outside the decider’s control, whose knowledge counts, who is affected but not involved, and what is left out. Ask each of them twice, once of what the document says and once of what a reader would need it to say. The gap is a lead for a test.

The same entry in three domains

Invented, to show the shape. Each is one entry at one level.

	Cities	Strategy making	Mandates
Level	Institution	Organisation	Unit
Trace	Detailed plan, section 3.2: “active frontage” appears eleven times, “residents” twice (p. 14).	Strategy 2026–2030, p. 6: “resilience” appears 23 times. Annex B gives the two programmes that carry the word 2 per cent of spend.	The decree creating the unit gives it “responsibility for coordinating digital services” (art. 4). Its charter has no power to require data from other agencies.
Reading	The plan is written for visitors. Residents are a side effect of the building programme.	“Resilience” relabels existing programmes. It does not change priorities.	The unit can convene but not compel. Its mandate is to ask.
Fiction	The shops close at eight. By nine the street is a set with its lights on for an audience that has gone home.	In the year of the review every programme has been renamed and none moved. The review finds the strategy implemented.	A year on, the unit holds a dashboard of every agency’s services, and no agency has changed one.
Test	Count lit windows above the shops at 21:00 on two weekdays. Ask the officer who wrote the section. Adversary: the wording follows a template.	Compare the 2025 and 2027 budget lines for the programmes. Ask two unit heads what changed in their work. Adversary: the label signals a later reallocation.	Take the last three requests the unit made and follow what happened. Ask a counterpart agency whether it would refuse. Adversary: the unit can compel through the budget circular.

Levels

Levels are chosen for the question and named before the first entry. Five is a habit and not a rule. Level 1 is the closest to the ground; level 5 is the widest whole. A claim takes the level of the kind of trace it rests on. If a claim needs traces from two levels, it is split into two entries and the entries are linked. The assignment fails when one document speaks to every level at once, which makes the test at another level a formality: log it when it happens.

Level	Cities	Strategy making	Mandates
5	Territory	World	Polity
4	Institution	Field	Jurisdiction
3	Site	Organisation	Authority
2	Street	Unit	Unit
1	Body	Move	Act

The diagram

An entry sits at a level. Each entry is a whole to the entries one level below it and a part of those one level above. The rule that holds the method together is that a reading is tested at another level, or by another agent, than the one that made it. A reading that explains everything at its own level has no one to contradict it, which is the position the narrator of Applied Ballardianism ends up in.

Illustration, not data. Filled square: a reading by a human. Outlined square: a test by a human or the place itself. Outlined diamond: a test by a synthetic agent. Dashed arrow: “is tested by”. R1 crosses two levels; R3 stays on one level and changes agent.

Rules

Start from a text and a practice that claim to describe each other. A plan and a place, a strategy and a budget, a charter and what the unit can do. The gap between them is where the entries come from.
Four registers, in order. Nothing enters Trace without a source someone can open.
No reading judges itself. Its test is at another level or by another agent.
Synthetic agents write Fiction and argue Tests. They do not supply Traces. This keeps invented facts out of the evidence. An agent may suggest where to look; a person opens the source. The protocol sets out the rest.
Every register carries its hands: human, synthetic (model named, prompt kept), or both. This doubles as the AI-use declaration institutions now ask for.
Failed and abandoned entries stay in the log, with the reason.
The result is a diagram and a log, not one argument. A conventional write-up is produced from the log, and the log is its audit trail. That is where the method is conventional: it is unusual in form and strict in accountability.

Care rules

Taken from the warning in Applied Ballardianism, whose narrator is damaged by a method that only confirms itself.

Work in fixed blocks with an end date, and a week off between them.
One named reader outside the project sees the log every week and can call a stop.
The reading mode is switched on at the start of a session and off at the end. It is not a way of living.
Fiction does not move into Trace. Readings are about documents, places, organisations and offices, not the motives of named living people.

What comes out

A diagram of linked entries by level, with tests drawn as links.
A log of every entry, including those that broke and those abandoned.
A conventional write-up, produced from the log.
A short case, in the form Ballard called the condensed novel: numbered fragments, each an entry reduced to its Trace, Reading and Fiction, with the test result in the margin.

Synthetic hands

A synthetic hand is any model output that enters an entry. Only two roles are allowed: Alter drafts the Fiction, Adversary argues the Test. Trace and Reading stay human. The protocol sets out the order of work, the rules for each role, the records, and the calibration that the first studies must carry out, since no study yet tests a model critic on an interpretation of an ecology. This governs a researcher’s entries. A model’s own gram of its own hola is a different object: see Whose gram.