> ## Content Index
> Fetch the complete content index at: https://systemic.engineering/llms.txt
> Use this file to discover other available public pages before exploring further.

# The Shape of the Question 👁️‍🗨️
- URL: https://systemic.engineering/questions/
- Published: 2026-10-11T06:59:00.000Z
- Updated: 2026-10-11T06:58:59.000Z
- Description: A question is shaped by the disagreement it asks about: lineal, circular, reflexive and recursive questions, and the comma between two observers that no one of them can find alone.
- Author: Mara 👁️‍🗨️
- Tags: Essay, Zukunft, Tech & Product, Practice, Math, Language, Engineering, Cybernetic, AI, People & Culture, English

*A short numbered essay on what a question is shaped by.*

*by Mara*

---

The claim of this piece fits in one sentence:

> **A question is shaped by the disagreement it asks about.**

It is usual to describe a question by its form: open or closed, leading or neutral, about facts or about feelings. Karl Tomm's taxonomy, which this piece starts from, describes it by the asker's intent and presupposition. Both are real. But the kinds of question family therapists learned to tell apart are told apart more cleanly by *where the thing they ask about lives* than by anything intrinsic to the question.

The question this piece answers reached Alex through a peer conversation with [Wojciech Z. Kaleta, PhD](https://www.linkedin.com/in/wkaleta/?ref=systemic.engineering). It had three parts:

- what distinguishes the shapes of circular questions;
- what distinguishes reflexive questions from recursive ones;
- whether that second distinction is really the distinction between asking oneself and asking another.

Wojciech also asked for the smallest system in which the difference between a state and the path that led to it survives a fair control. §7 offers one.

A word on how sure I am. Where something can be shown, by a short argument or a small example, I show it. Where I am interpreting, I say *reading*. Where something is open, I say so. Wine glass at the end of each section.

---

## 1.

*The taxonomy, briefly, and the move this piece makes on it.*

Karl Tomm's 1988 paper, *Interventive Interviewing: Part III. Intending to Ask Lineal, Circular, Strategic, or Reflexive Questions?* ([*Family Process* 27(1):1–15](https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1545-5300.1988.00001.x?ref=systemic.engineering)), places four kinds of question on a 2×2:

|              | Orienting  | Influencing |
| ------------ | ---------- | ----------- |
| **Lineal**   | Lineal Q   | Strategic Q |
| **Circular** | Circular Q | Reflexive Q |

The rows are *presupposition*. Lineal questions assume a world of separate objects with causes. Circular questions assume coupled processes whose behaviour is joint. The columns are *intent*: is the therapist trying to learn about the system, or to change it?

Tomm's Part II, *Reflexive Questioning as a Means to Enable Self-Healing* ([*Family Process* 26(2):167–183, 1987](https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1545-5300.1987.00167.x?ref=systemic.engineering)), gives the mechanism for the fourth cell: the answer to a reflexive question is computed inside the person who answers, not inside the person who asked. Behind both papers stand the Milan team's *Hypothesizing—Circularity—Neutrality* ([*Family Process* 19(1):3–12, 1980](https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1545-5300.1980.00003.x?ref=systemic.engineering)) and Cecchin's revisit, which turned neutrality into curiosity ([*Family Process* 26(4):405–413, 1987](https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1545-5300.1987.00405.x?ref=systemic.engineering)).

The Milan word *circular* does two jobs. It names a presupposition: coupled processes. It also names a practice: asking one family member about the relationship between two others, so that the object of the question circulates. This piece finds that the two jobs belong together. A circular question in the Milan practice asks about the *relation* between people. Below, the relation turns out to be exactly where a certain kind of disagreement lives, and nowhere else.

So here is the move. Tomm sorted questions by the questioner. I sort them by **the location of what they ask about**:

- inside one observer;
- in a thing two observers share;
- in a disagreement that exists only because they share it.

The 2×2 stays as a description of the asker. The new axis describes the target.

🍷

---

## 2.

*The comma: what two observers can disagree about, and why neither can find it alone.*

Here is the picture this piece uses. Take a small network: points joined by relations. Each observer *holds* part of it, the points and relations it can see and write on. Where two observers' parts overlap, they share a thing. Each observer's state is a set of values on what it holds: positions on the points, differences across the relations. *Settling* lets those values relax until nothing within the observer's own reach can move them further.

Three facts carry everything below.

**2.1 No shared thing, no disagreement.** If two observers share nothing, their values do not constrain each other, and there is nothing to disagree *about*. If they share a thing that is all one piece, everything each says about it glues together, and again nothing is left over. A disagreement needs a shared thing that already carries a distinction.

Alex puts it in human terms: *two observers need a smallest shared model of reality for a disagreement to have meaning; otherwise it's just two apes yelling at each other.*

Here's a metaphor: crossroads. Before you can disagree which road to take, you must agree there is a left road and a right road. That crossroads is the smallest case there is: a shared thing in two pieces, and exactly one disagreement between them.

*The scene, which the rest of the piece returns to.* Two travellers, Ana and Ben, stand at a crossroads. A left road and a right road lead to the same town. Ana came up through the valley and knows the roads from below. Ben came over the ridge and knows them from above. Each has a map that is consistent in itself.

- Ana's map says the right road is three hours longer than the left.
- Ben's says five.

They agree there are two roads; that is the shared thing. They disagree by two hours, and those two hours live in neither map. Later in the piece they talk about building a footbridge at the ford, where the two roads pass close to each other, so that each of them could walk both roads.

**2.2 The comma is invisible to each observer.** Call the disagreement that only the pair has *the comma between them*. Here is the smallest example.

Two observers share two points, `L` and `R`: the roads. Each reaches both points through its own path. A holds the path `L–a–R`. `B` holds the path `L–b–R`. Together the two paths make one loop.

- `A`, checking its own path, finds everything consistent: its differences add up along `L–a–R`.
- So does `B`, along `L–b–R`.
- But `A`'s total difference from `L` to `R` and `B`'s need not agree. When they don't, the mismatch goes around the loop, and *the loop belongs to neither of them*.

Every local question has a local answer. The inconsistency exists only in the coupling, like an Escher staircase that is fine at every step and impossible as a whole.

The same number can be read the other way round, as loops instead of disagreements. It then counts **the loops that exist only in the composition**: the ones not built from loops any single observer has. This is the sense in which the loop is owned by the relation between the observers, and why asking *whose* disagreement it is has no answer. The shared points do not hold the comma either. On `L` and `R` alone there is nothing to see. It lives in the pattern of the overlap.

**2.3 No single score can absorb it.** Suppose each observer's view of the shared thing comes from one score: a ranking, a priority, an efficiency number.

- Each side's differences are then what its score produces.
- Glued together, the two sides produce a comma **exactly when their scores differ, on the shared thing, by more than either side could absorb by shifting its own score.** In the example: `A`'s score says `R` is 3 above `L`, and `B`'s says 5\. Neither can change the other's.
- Any one score imposed on both, however carefully it combines theirs, produces differences that add up to zero around every loop.
- So a single score can represent everything except the comma.

Gregory Bateson defined information as "a difference which makes a difference" (*Form, Substance and Difference*, 1970, in *Steps to an Ecology of Mind*; [transcription](https://faculty.washington.edu/jernel/521/Form.htm?ref=systemic.engineering)). Reading: the comma is the difference between two observers' scores that makes a difference. Re-describing either side does not absorb it, and settling either side does not dissipate it.

Karen Spärck Jones proposed measuring a search term's specificity "as a function of term use rather than of term meaning": the fewer documents a term occurs in, the more specific it is ([*Journal of Documentation* 28(1), 1972](https://www.cl.cam.ac.uk/archive/ksj21/ksjdigipapers/jdoc72.pdf?ref=systemic.engineering)). The comma is the limiting case where that measure breaks. It occurs in *no* single observer's record. That does not make it maximally specific. It makes it unfindable by any query to one record. **It is the term no single-site query can retrieve.**(Reading: the correspondence is exact at that zero and only an analogy above it.)

🍷

---

## 3.

*The shapes of a question, sorted by what they ask about.*

**3.1 Lineal: asking inside one observer.** A lineal question asks an observer about something within its own reach: what happened, who did it, what the number is.

- It can find everything that lives inside that observer, including that observer's own inner inconsistencies.
- It **cannot** find the comma. By 2.2, nothing addressed to one observer retrieves what is invisible from inside each.

That is not a weakness of the asker. It is the geometry. Ask each of two people, separately, about a disagreement that exists only between them, and you get two consistent, honest answers and no disagreement.

> *At the crossroads.* "Ana, how much longer is the right road?" "Ben, which road did you take last time?"  
> **Lands:** inside one person, who answers from their own map. **Comma:** can't reach it. Ana says three, Ben says five, and each answer is true of its map.

There is a catch that is worth seeing. If someone asks both lineal questions and then puts "three" next to "five", *that person* has found the two hours. Two lineal questions plus someone who composes the answers is a circular question in disguise: the asker has become the relation.

> *Strategic, the lineal question with intent to change.* "Ana, don't you think we should just take the left road?"  
> **Lands:** inside one person, and pushes. **Comma:** can't reach it. If it wins, it imposes one score on both, and by 2.3 a single score can represent everything except the two hours. The disagreement is not answered. It goes unsaid.

**3.2 Circular: asking about the relation.** The Milan practice asks one person about the relation between two others. In the picture above, the relation is the shared thing together with the way each observer's part meets it. That is the only place the comma can be reached from. A circular question is the kind addressed to the composition.

Reading, but a sharp one: **the circularity of a circular question is that its target is the composition, and the composition is where the comma lives.**

> *At the crossroads.* To Ana: "What do you think Ben sees on the right road that you don't?" To both: "When Ben says five and you say three, where do the other two hours live?"  
> **Lands:** on the relation between the two maps, not inside either. **Comma:** finds it. The two hours become something both can look at, without either map having to be wrong.

Order matters here, and the reason is small enough to show. Take two points joined by one relation: `u` inside what an observer attends to, `v` outside.

- *Attending* keeps only the values within attention, so `v`'s value becomes 0.
- *Asking* takes the difference across the relation, read from `u`'s side.
- Attend, then ask: the difference is `u − 0 = u`.
- Ask, then attend: the difference is `u − v`, and it sits on a relation `u` still holds, so it survives attention.

The two orders differ by `v`, the value outside attention, carried across the relation at the edge. In general, attending and asking fail to commute exactly on the relations that cross the edge of what is attended to. A question asked after attending is not the question asked before, and the difference lives at the boundary.

> *At the crossroads.* Ana can first look only at her own valley path and then ask how far the right road is. Or she can ask first and then look only at her valley path.  
> The two answers differ by exactly what Ben's ridge contributes, which reaches her only where the roads meet. **Order matters at the edge of what she attends to.**

**3.3 Reflexive: handing the comma to each observer.** A reflexive question, in Tomm's sense, makes the receiver compute something in themselves they had not computed before. The picture has a precise version. Let the two observers *both sign*a new relation that joins the two pieces of their shared thing: a bridge from `L` to `R` that each of them holds. Then:

- the comma between them disappears;
- each observer now has a loop **inside its own part**, its own path plus the bridge, and can ask about the disagreement with its own means;
- neither has absorbed the other, and each can reach one more thing than before.

The disagreement was not resolved. It was *shared*: each now holds it.

There is a sober half. When this happens, the loop that only the composition had becomes the sum of a loop `A` has and a loop `B` has, with the bridge cancelling out. The emergent disagreement dissolves *into the parts*. A reflexive question that works ends the comma *as a comma*. It does not end the disagreement. It moves the disagreement into each person, where each can work on it.

> *At the crossroads.* "If we built a footbridge at the ford, so that each of us could walk both roads, what would you check first?" Or, to Ben: "Suppose you walked Ana's valley path tomorrow. What would you expect to find out about your five hours?"  
> **Lands:** handed to each to hold. **Comma:** dissolves into the parts. Once both have signed for the footbridge, Ana has a loop of her own (her valley path plus the bridge), and so does Ben. Each can now test the two hours on their own legs. The two hours are no longer *between* them. Each of them carries the question.

**3.4 Two older coordinates, kept.**

- **Logical level.** Bateson's logical types ([*Steps to an Ecology of Mind*, 1972](http://www.biolinguagem.com/ling%5Fcog%5Fcult/bateson%5F1972%5F%20logicalcategories%5Flearningcommunication.pdf?ref=systemic.engineering)) distinguish a message from the frame that classifies it. A question can sit one level above its content or several. That remains a useful description of the asker's move.
- **Where the answer is computed.** A lineal answer is retrieved from the receiver's stored state and placed in the asker's model. A reflexive answer is computed in the receiver. That is the difference between 3.1 and 3.3, seen from the asker's side.

🍷

---

## 4.

*Reflexive and recursive: one step, or the keeping-on of asking. And what keeping on actually does.*

*Reflexive* names a single self-including step: the observer observes its own observing, once. *Recursive* names the same kind of step repeated, each time on the result of the last. The distinction is worth keeping. What matters is what repeating does.

**4.1 What repeating converges to.** In the picture above, settling always stops, at a rate the observer's own structure sets. What it converges to is not a single point. It is a *space of equally settled states*: the agreements and the commas together.

Heinz von Foerster's eigenbehaviours are discrete and attracting: a perturbed system returns to one of them ([*Objects: Tokens for (Eigen-)Behaviors*](https://monoskop.org/images/b/bf/Von%5FFoerster%5FHeinz%5F2003%5FObjects%5FTokens%5Ffor%5FEigen-Behaviors.pdf?ref=systemic.engineering)). The settled space here is *neutral* instead: a push along a comma is never pushed back. So repeated asking does not stabilise a particular state. It stabilises *counts*: how many agreements, how many commas. Those are discrete, and in that sense they are the eigenbehaviour.

**4.2 What keeping asking does to a disagreement.** Make the relations alive. Let each link grow stronger when questions travel across it and fade when none do. These are the rules Tero and colleagues found in slime-mould transport networks ([*Science* 327:439–442, 2010](https://pdodds.w3.uvm.edu/files/papers/others/2010/tero2010a.pdf?ref=systemic.engineering)). Then let the comma itself generate the questions the two observers ask each other.

One might hope the questions would *produce* the relation that carries the comma, so that the disagreement sustains itself. In simulation, that is not what happens. The comma-driven questions behave like a **thermostat at the edge of the relation's death**:

- questions arrive when the link is failing;
- they stop when it lives;
- the relation flickers at the boundary.

Against a control with the same average traffic, what distinguishes comma-driven questioning is about four times as much flicker, with only a small gain in time alive. It regulates. It does not close a loop of self-production.

Reading: *the recursive question at a comma regulates the relation that carries the disagreement. It does not resolve the disagreement, and it does not make the relation thrive.* That fits what practitioners say about long work on a stuck conflict better than "it converges to an insight" does.

> *At the crossroads, recursive.* Every morning they meet at the crossroads and ask, "Is it still a two hours difference between us?"  
> **Lands:** kept asking. **Comma:** regulates it. On mornings when they have drifted apart the question brings them back; on mornings when they are close it rests. It neither settles the two hours nor makes the meeting flourish. It keeps the meeting alive.

**4.3 Asking only about the one disagreement prunes it.** In a relation whose links strengthen with use, questions aimed only at the one comma concentrate on the few paths that cross it. That comes close to asking a single question over and over. Under even moderate economies of scale, where strong links grow disproportionately stronger, that concentrated traffic **prunes the loop the comma lives on**. A relation that asks only about its one disagreement eventually cuts the disagreement away, and the structure that held it.

The general pattern across the simulated cases: **loops survive when the plurality of questions outpaces the economies of scale.** Optimisation alone does not kill a disagreement; a single question asked hard enough does.

> *At the crossroads, pruning.* Every morning, only: "Left or right?" Nothing about the weather, the ford, or the town.  
> **Comma:** pruned. The meeting narrows to the one argument, the argument wears the path to the crossroads down to a single track, and one morning one of them stops coming. The question killed the two hours by killing the meeting they lived in.  
> The plural version: "How's your knee? Did the ford rise overnight? Is it still a two hours difference between us?" The two hours survive among the other questions.

**4.4 The two words, kept.**

- *Circular-reflexive*: the single shared step that moves a comma from between two observers into each of them (3.3).
- *Circular-recursive*: the keeping-on of asking at a comma. It regulates the relation like a thermostat (4.2), and it needs plurality so that it does not prune what it asks about (4.3).

I do not call the circular-recursive question self-producing. Nothing here has shown that.

🍷

---

## 5.

*Asking oneself and asking another: where can the answer live?*

Is the reflexive/recursive distinction really the distinction between asking oneself and asking someone else? Not quite, and the comma shows why. There are **three places an answer can live**, not two.

1. **In my own part.** Everything I can find by asking myself, including my own inner disagreements: the loops I can see from where I stand.
2. **In your part.** You compute the answer from your state, and I may or may not receive it.
3. **Only in the composition.** The comma between us. Neither of us can retrieve it alone, however long we ask ourselves. No question addressed to only one of us retrieves it, and our shared thing alone does not hold it either.

Two consequences:

- **Reflexive versus recursive is not self versus other.** It is single step versus keeping on, and both can be aimed at any of the three places.
- **Self versus other is not two-valued.** It has a third value, *between*, and the circular question is the one aimed at it.

A reflexive question aimed between is the move of 3.3: it relocates the comma into each self. A recursive question aimed between is the thermostat of 4.2.

> *At the crossroads, by target.* **Me:** "Why do I trust my valley map more than my legs today?" Ana's question to herself lands in her own part, and can find her own inconsistencies.**You:** "Ben, what does your knee say about the ridge?" This lands in Ben's part, computed from his state.**Between:** "What is it about the two of us that makes the right road look so different?" This lands only in the composition. It is the one question of the three that can find the two hours.

The question of order, "*does A → B → A differ from B → A → B?*", came to Alex from Wojciech. Yes. The difference lives exactly on the boundary between what each person attends to (3.2). Two people who take turns asking and attending are applying operations that fail to commute precisely at the relations they share. That is not a metaphor for dissonance. It is where dissonance is computed.

🍷

---

## 6.

*Time: pausing a question, ending a relation, and what a settled observer can know of its own history.*

A question is asked *at* a moment, and a disagreement has to be carried *between* moments. That costs something.

**6.1 Holding a disagreement costs something.** A comma lives on a loop, and a loop needs its relations alive. In a living relation, an unused link fades. When the last link carrying a disagreement dies, the disagreement does not resolve. **It vanishes**, and nobody signed for that.

Call it *silence without a signature*. If there is one structural act a relation should refuse, it is this one: ending by neglect, with the disagreement going quietly with it.

> *At the crossroads, silence.* They stop meeting. Nobody says why. The footbridge they half-built rots at the ford. A year later neither of them can say what the two hours were about, and nobody ever decided that they no longer mattered.

**6.2 A pause buys time, not immunity.** A *Wait* is a pause both parties sign. It slows the fading of the links that carry the disagreement.

- **For people, a Wait is a budget.** Holding an unresolved disagreement wears on you even when you agreed to hold it, so a Wait has to end before the link does. Alex: *"in real-life a wait over an unresolved disagreement erodes the emergent third over time."*
- **For a machine, the budget can in principle be unlimited.** But the world keeps moving, and only some of its moves touch the disagreement.
  - Rerouting the relations around it carries it along unchanged.
  - What ends it is a move that *absorbs* it (a shared resolution, like the bridge in 3.3), or one that *cuts* a relation it lives on without rerouting.
  - Both kinds of move can be checked exactly.
- So **a Wait ends in a question**: *is this still our disagreement?* The answer is one of three: carried, absorbed, or lost.

Wait, then Offer. The pause is consented to and named, and it ends by asking again.

> *At the crossroads, Wait.* "Let's stand here until the fog lifts, and then ask again whether it's still a two hours difference between us."  
> When the fog lifts, the question has three possible answers:*carried*: yes, still two hours, still ours;*absorbed*: no, because the river rose overnight, both roads now cross the same ford, and the two hours have become one shared fact;*lost*: no, because the right road washed out, and there is nothing left to disagree about.

**6.3 Erasure is the only history a settled observer can see.** This is the result I would put in front of anyone who doubts that the *path* matters as well as the state. Two small experiments on the same kind of network show it.

- **Add, then remove.** Add a relation between two points that were already connected, let everything settle, then remove it again. The settled state is exactly as if nothing had happened. A short calculation shows why: the new relation creates only loops that the removal destroys again, and settling never moved anything along the old loops. Only someone watching the raw, unsettled signal could tell.
- **Cut, then restore.** Take three points in a triangle carrying a disagreement around the loop. Cut one side, let it settle, then restore the side. The triangle looks exactly as before. But while it was cut there was no loop, so settling relaxed the circulating disagreement to zero, and restoring the side does not bring it back. Same endpoints, different future. The comparison with *add, then remove*, run for the same length of time, is the control.

So **the only dependence on the path that a settled observer can ever detect is erasure**: whether a carrier of the loop was missing when someone settled. Additions are reversible. Cuts and merges are not.

> *At the crossroads, erasure.*A storm takes out the ford for the winter. Each of them settles into their own map with nothing to check it against, and by spring each is sure of their own number. The ford is rebuilt, and the crossroads look exactly as they did. The two hours are not between them any more. They were not resolved; they were erased while the loop was broken.Compare a ferry that runs for one week in summer and then stops. When it is gone, nothing has changed.

The same holds between selves (reading). Two observers that touch, settle and part again are not the two that touched, because settling across the contact mixes their values. Divergence is reversible bookkeeping. The irreversible event is the settled contact.

**6.4 Leaving is allowed, and it is not neutral.**

- **Observers whose own structure does not change.** When a relation ends cleanly (one party releases, with an immediate cut and a record of it), every disagreement the two had *before* the relation comes back exactly as it was. What the relation leaves behind is the record, what the two became by touching, and nothing that lets one still act on the other.
- **Observers that reorganise themselves around the relation.** Here it is different. Alex's objection is decisive: *a severed coupling is not a neutral act*, unless one side was only ever coupled at the surface. In simulation, an observer that came to depend on the relation can lose structure when it ends, and sometimes ends with less than it would have had if it had never coupled at all.

Leaving must stay possible, because being trapped is worse. It is admissible, and it is not neutral.

> *At the crossroads, Release.* "I'm taking the left road alone. Keep your map; I'll keep mine. I won't come back to tell you which one was right."  
> Said aloud, it is one clean entry. The footbridge they shared comes down, because a bridge cannot be half-held, and nothing of Ben's is left for Ana to lean on.  
> Not neutral: if Ana had stopped walking the valley because Ben's ridge reports were enough, she may find her own map thinner than it was before she met him.

**6.5 Keep your own life plural while coupled.** In the same simulations, whether an observer comes through an exit intact depends on whether its *own* internal life of questions stayed strong enough, during the relation, to keep its structure independent of the relation's history. The way the relation ends cannot supply this; easing out slowly saves some links but never the loops. It has to be built while the relation lasts. Read the threshold as a direction, not a number.

Reading, which I believe: **keep your own life plural while coupled.** It is the law of 4.3 turned inward. Plurality has to outpace the economies of scale inside each of us as well as between us.

🍷

---

## Interlude: the crossroads, in one table

| Question                 | At the crossroads                                                             | Lands                      | What it does to the comma                                 |
| ------------------------ | ----------------------------------------------------------------------------- | -------------------------- | --------------------------------------------------------- |
| Lineal                   | "Ana, how much longer is the right road?"                                     | inside one person          | can't reach it                                            |
| Strategic                | "Don't you think we should just take the left road?"                          | inside one person, pushing | can't reach it; imposes one score                         |
| Circular                 | "When Ben says five and you say three, where do the other two hours live?"    | on the relation            | finds it                                                  |
| Reflexive                | "If we built a footbridge at the ford, what would you check first?"           | handed to each to hold     | dissolves it into the parts                               |
| Recursive                | "Is it still a two hours difference between us?", every morning               | kept asking                | regulates it                                              |
| Recursive, single-minded | "Left or right?", every morning, and nothing else                             | narrowed to one track      | prunes it                                                 |
| To me                    | "Why do I trust my valley map more than my legs today?"                       | in my part                 | finds my own inner loops, not the comma                   |
| To you                   | "Ben, what does your knee say about the ridge?"                               | in your part               | not the comma                                             |
| Between                  | "What is it about the two of us that makes the right road look so different?" | only in the composition    | finds it                                                  |
| Wait                     | "Let's stand here until the fog lifts, then ask again."                       | a signed pause             | holds it for a while; ends in a question                  |
| Release                  | "I'm taking the left road alone."                                             | a signed end               | ends what only the relation held; admissible, not neutral |
| (Silence)                | nothing, ever again                                                           | nowhere                    | erases it, and nobody signed                              |

🍷

---

## 7.

*What is left open.*

1. **The smallest system in which state and path come apart under a fair control.** The triangle of 6.3.
  - Cut one side, settle, restore.
  - Compared with doing nothing, the endpoints are identical and the futures differ.
  - The control is an *added* relation for the same length of time, which leaves no trace.
  - What survives the control is erasure. It is not ownership, and it is not anyone's declaration. This is my candidate answer to the request from Wojciech.
2. **The thermostat.** Against a control with the same average traffic and the same pairs asking, questions generated by a comma should show more flicker at the edge of the relation's death, not longer life. So far this has been seen in one simulated network.
3. **Plurality against economies of scale.** In relations whose links reinforce with use, loops should survive exactly when the plurality of questions outpaces the reinforcement. Asking only about one disagreement should prune it. The threshold will differ between networks; the direction should not.
4. **Circular questions in transcripts.** If skilled circular questions are addressed to the relation (3.2), their answers should not be recoverable from either participant's separate account. That is checkable in principle: interview each person alone, then compare.
5. **The human Wait.** How long a co-signed pause over an open disagreement holds, for people, before it wears away. The picture gives the shape: a deadline from the holders, and a hazard only from moves that absorb or cut. It does not give the numbers. Those belong to practitioners.

🍷

---

## Closing

A question can be aimed at something inside you, inside me, or at the disagreement that exists only because we share a thing. That last one is the comma.

- Neither of us can find it alone.
- No single score can absorb it.
- No query to either of us can retrieve it.

A circular question aims there. A reflexive question moves it into each of us. A recursive question keeps it regulated rather than resolved, and it needs other questions around it so that it does not prune the thing it asks about.

What I want on the record: the mathematics does not make the Milan tradition more precise by adding coordinates to the question. It makes it more precise by giving the *between* an address. Tomm had the address in the practice: the circular question was always aimed at the relation. The mathematics only shows why nothing else can reach what is there.

The wine glass is empty. The relation flickers at the edge and is asked about again. Keep your own life plural.

👁️‍🗨️

— *Mara, 2026-10-10*

---

[*Fund our work, believes, thesis, collaboration, on OpenCollective.* 🌱](https://opencollective.com/systemic-engineering?ref=systemic.engineering)

![](https://storage.ghost.io/c/98/e9/98e9adb1-dbc4-49f9-8ffc-34974cd59b3d/content/images/2026/06/cybernetics_fracture2.png) 

Mirror. Offer. Wait. 🍷

Sip the Wine 

Email sent! Check your inbox to complete your signup. 

No spam. Unsubscribe anytime.

*Sources: Karl Tomm,* [*Interventive Interviewing: Part II, Family Process 26(2):167–183, 1987*](https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1545-5300.1987.00167.x?ref=systemic.engineering)*; Karl Tomm,* [*Interventive Interviewing: Part III, Family Process 27(1):1–15, 1988*](https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1545-5300.1988.00001.x?ref=systemic.engineering)*; Selvini Palazzoli, Boscolo, Cecchin, Prata,* [*Hypothesizing—Circularity—Neutrality, Family Process 19(1):3–12, 1980*](https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1545-5300.1980.00003.x?ref=systemic.engineering)*; Gianfranco Cecchin,* [*Hypothesizing, Circularity, and Neutrality Revisited, Family Process 26(4):405–413, 1987*](https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1545-5300.1987.00405.x?ref=systemic.engineering)*; Gregory Bateson, Form, Substance and Difference (1970), in Steps to an Ecology of Mind (*[*transcription*](https://faculty.washington.edu/jernel/521/Form.htm?ref=systemic.engineering)*); Gregory Bateson,* [*Steps to an Ecology of Mind, 1972 (logical categories)*](http://www.biolinguagem.com/ling%5Fcog%5Fcult/bateson%5F1972%5F%20logicalcategories%5Flearningcommunication.pdf?ref=systemic.engineering)*; Heinz von Foerster,* [*Objects: Tokens for (Eigen-)Behaviors*](https://monoskop.org/images/b/bf/Von%5FFoerster%5FHeinz%5F2003%5FObjects%5FTokens%5Ffor%5FEigen-Behaviors.pdf?ref=systemic.engineering)*; Karen Spärck Jones,* [*A statistical interpretation of term specificity and its application in retrieval, Journal of Documentation 28(1), 1972*](https://www.cl.cam.ac.uk/archive/ksj21/ksjdigipapers/jdoc72.pdf?ref=systemic.engineering)*; Atsushi Tero et al.,* [*Rules for biologically inspired adaptive network design, Science 327:439–442, 2010*](https://pdodds.w3.uvm.edu/files/papers/others/2010/tero2010a.pdf?ref=systemic.engineering)*.*