Finite encounters
Every interaction a system could have was written in advance and reviewed by someone who built it.
A capability gap emerging from our work · 2026
Across our work on machine-assisted inquiry, participation, coordination and agreement, a new design need is becoming visible.
The emerging practice concerns how a computational system enters a human situation—and how it earns the right to remain there.
[ what has changed ]
Until recently every screen, message and flow passed through a designer’s hands before it reached a person. Generative systems have ended that condition.
Every interaction a system could have was written in advance and reviewed by someone who built it.
A machine counterpart composes its side of an interaction at the moment of use—in situations no designer anticipated and most will never review.
When the instances cannot be authored, the generator must be.
[ the design question ]
A system must learn how to enter a human situation appropriately.
[ a live contingency field ]
Change the human situation below. The purpose remains stable while the prompt, tone, timing and permission structure are recomposed.
[ two orders of capability ]
The two levels work together. Interaction design shapes each encounter; meta-design establishes the conditions under which that encounter becomes appropriate.
Shapes a particular encounter between a person and a system: what appears, what can be done, how an action is understood and how the system responds.
Designs the syntax through which interactions are generated, selected, connected and revised across people, machines, institutions and time.
The unit of meta-design is a field of possible interactions—and the conditional logic through which a system moves within it.
[ the wager · the cost of attentiveness ]
No rule fully captures a situation—meta-design accepts this entirely. And attentiveness has a price: a system that adapts to your fatigue has been measuring your fatigue. Sensing is a governed resource, never a free good.
Prefer the context people give to the context that must be read.
Visible, callable, contestable—retired when its warrant lapses.
Collect the situation the move needs, not the situation the system could use.
A system that changes register can say that it has, and why.
Misreading made visible, cheap to correct, and incapable of quietly compounding.
[ interaction-generating grammar ]
Meta-design holds a sequence of contingent choices. It connects the situation to the form of an intervention, then returns the consequences as material for revision.
Context · history · role · power · attention · risk
Syntax · tone · timing · modality · permission · threshold
Ask · show · wait · connect · propose · escalate
Response · consequence · adaptation · contest · repair
[ response architecture ]
In personalisation the person is the target of the adaptation—tuned around them, invisible to them. In meta-design the person is a party to it: inferences declared, the adaptation legible, and qualification, redirection and refusal always open.
Continue with informed consent.
Preserve ambiguity, condition and nuance.
Ask for explanation or grounds.
Close the exchange without concealed penalty.
Retain state and return when capacity changes.
Open an inference or decision to review.
Reopen an earlier position or commitment.
Pass the situation to a person or collective.
The grammar may adapt anything except the means by which people contest the grammar.
[ negotiated co-presence ]
Legitimacy is a functional requirement, not an ethical garnish: these systems exist to hold coordination among many actors, across long durations, in problem spaces no single actor owns. The machine’s role must remain invited, bounded, intelligible, interruptible and revisable.
Renewed as the people, conditions and consequences of the situation change.
Why is the system entering now?
Who gave it the right to ask?
What is it inferring or recording?
Can participation be paused or reshaped?
What does the participant receive?
How can an inference be contested?
Whose experience is absent?
Who carries the effects?
[ an evolving grammar ]
Every encounter produces evidence about fit, exclusion, misunderstanding and consequence. Memory should deepen situated judgement while preserving difference, refusal and surprise.
The invitation arrived when attention was already exhausted.
The system offered a pause before asking for a response.
Ask only when the person has chosen a suitable moment.
[ the evidence ]
This is not a speculative requirement. In our own systems the work kept falling between the professions—so we assembled the discipline ad hoc, by hand.
The invitation that opens a morning’s thinking becomes pressure at midnight. What resolves the difference is not two prompts but one rule—and someone had to write it.
A first generative grammar for what a machine may say to a person: the warrant for speaking now argued inside the ask, and silence as the designed default.
How a changing field of commitments responds when one participant withdraws.
The work fell between interaction design, conversation design and systems engineering—and belonged fully to none of them.
The assembly is the evidence: the gap is real, because we keep filling it by hand.
[ the capability ]
Someone able to move between semantics, pragmatics, visual hierarchy, conversational rhythm and parametric system behaviour—between a sentence and the system that generated it. Its deliverable is concrete: a written grammar, inspectable and revisable, so the capability can be held by a team as well as embodied in a person.
Spacetime-contingent thinking across language, systems and relations.
Syntax, pragmatics, register and implication.
Framing, sequence, tone and attention.
Hierarchy, rhythm, legibility and the shape of an encounter.
State, rules, parameters, memory and feedback.
Consent, power, legitimacy and repair.
Time, place, relationship, culture, institution and disposition.
[ an invitation ]
“This may be a field we now need to build together.”
This practice may already exist—in an individual whose work crosses linguistics, communication, visual design, computational systems and situated judgement; in a team that has had to invent it mid-project; in a practice working under another name. Whatever you currently call it, this is a capability gap already being exposed by our work.
If you have been developing this intersectional practice, please get in touch with us.
↗A note on this pageIn the spirit of honesty: this piece would itself be much better with a meta-designer. It meets every reader in the same form, at the same register, at whatever hour they arrive—and it cannot hear what they would say back. It was designed. It was not yet meta-designed.
Capability note · 2026
A Capability Gap Emerging from Our Work
Across our work on machine-assisted inquiry, participation, coordination and agreement, a new design need is becoming visible.
We are increasingly building systems in which a machine may display information, execute an instruction, ask someone a question, surface tacit knowledge, test an assumption, connect an individual contribution to a shared mission, make a changing field of commitments visible, or support several people in reaching a decision. The design challenge now encompasses the clarity of the interface, the coherence of the visual artefact and the appropriateness with which the system enters a human situation.
It has long been understood that the meaning of an interaction depends on its situation. What is new is the scale on which that insight now operates. Until recently, the interactions a system could have were finite and authored: every screen, every message, every flow passed through a designer’s hands before it reached a person. Generative systems have ended that condition. A machine counterpart now composes its side of an interaction at the moment of use, in situations no designer anticipated and no designer will ever review. Most of the encounters such a system will have were never seen by anyone who built it.
This is the fact that forces design up a level. When the instances cannot be authored, the generator must be. The designer’s material is no longer the encounter but the conditions under which encounters are composed.
That appropriateness is contingent. The meaning of a prompt changes according to when it is asked, what has preceded it, who is asking, the relationship of power involved, the person’s available attention and the consequences of their response.
A question asked in the morning carries a different quality when asked late at night. A direct formulation may support action in an emergency and exert pressure in a reflective process. Tone, rhythm, sequence, visual density, silence, hesitation and refusal are all part of the situation.
This points beyond the design of a single optimal prompt, interface or journey. The task is to design the grammar through which appropriate interactions can be generated in relation to a particular person, purpose, place and moment.
The task is to design the grammar through which appropriate interactions can be generated.
In this framing, interaction design provides the essential first-order capability. It shapes a particular encounter, or sequence of encounters, between a person and a system: what appears, what can be done, how an action is understood and how the system responds.
Meta-design introduces a second-order capability. It designs the syntax through which multiple interactions are generated, selected, connected and revised across people, machines, institutions and time. Design attention expands from the quality of each encounter to the relationships among encounters: how one person’s contribution changes what another is asked; how an earlier refusal reshapes subsequent machine initiative; how permission, confidence and uncertainty accumulate or decay; and how the system changes register as the purpose, relationship, place or moment changes.
The two levels work together. An interaction designer may shape the form and experience of a question. A meta-designer establishes the conditions under which that question becomes appropriate: who may ask it, what knowledge authorises it, how tentatively it should be phrased, what its answer is permitted to change, how it relates to previous and future questions, and when silence is the appropriate response.
The unit of meta-design is therefore a field of possible interactions and the conditional logic through which a system moves within it. Meta-design creates an interaction-generating grammar.
There is a long and serious tradition of research which holds that no rule can fully capture a situation — that plans and rules always underdetermine what a moment actually requires. Meta-design accepts this finding entirely. Its wager is precise: the gap between rule and situation cannot be closed, but it can be governed. A system can hold every one of its models as a partial interpretation of a human situation; it can say so; it can phrase its moves tentatively, declare what it has inferred, ask before it acts, accept refusal without penalty, and treat silence as its default rather than its failure.
The discipline of meta-design is not the elimination of misreading. It is the construction of relationships in which misreading is visible, cheap to correct and incapable of quietly compounding.
The gap between rule and situation cannot be closed — but it can be governed.
Attentiveness has a price, and the argument must meet it directly. To enter a situation appropriately, a system must read the situation: the timing, the attention available, the power in the room, the consequences of a response. A system that adapts to a person’s fatigue has, by some means, been measuring their fatigue. The sensing that appropriateness requires is the same activity that, ungoverned, erodes the legitimacy appropriateness is meant to serve.
Meta-design therefore treats sensing as a governed resource rather than a free good. It prefers context that people declare over context that must be inferred. It treats every inference as a debt: visible to the person it concerns, callable and contestable by them, and retired when its warrant lapses. It senses minimally — collecting the situation it needs for the move at hand, not the situation it could use. And it makes the adaptation itself part of the interaction: a system that has changed register should be able to say that it has, and why.
Conventional personalisation offers a useful point of reference. It generally adapts an experience towards a predetermined outcome: greater attention, engagement, conversion or compliance. But the deeper difference is structural, not tonal. In personalisation the person is the target of the adaptation: it is tuned around them and invisible to them. In meta-design the person is a party to it: the adaptation is legible, its inferences are declared, and it can be qualified, redirected or refused.
Meta-design orients adaptation towards the quality and legitimacy of participation. It creates machine-mediated environments that can work through partial and situated knowledge, communicate the limits of their inferences, use tentative language, ask permission, preserve the capacity to withdraw, and allow people to qualify or revise what the system believes it has understood.
Legitimacy is not an ethical preference layered onto this practice; it is a functional requirement of what these systems are for. The systems we are describing exist to hold inquiry, coordination and agreement among many actors, across long durations, in problem spaces no single actor owns or can command. Participation of that kind cannot be extracted or purchased at scale; it survives only where it is trusted, contestable and fairly held — and it collapses, quietly and then completely, where it is not. A machine-mediated environment that optimises engagement while degrading legitimacy is not underperforming on a soft metric. It is consuming the resource it runs on.
The design objective is therefore to increase a person’s capacity to reflect, judge, contribute and act within a shared situation.
The response architecture becomes the central object of design. It governs how language, visual form, timing, sequence, permissions and degrees of machine initiative change across contexts. It combines invariant commitments with parameters that can adapt as the situation changes — and the boundary between the two is itself principled: the grammar may adapt anything except the means by which people contest the grammar. Rights, provenance, the visibility of inference and the capacity for refusal are fixed, because they are the conditions under which everything else can be safely adaptive.
This is the emerging field of meta-design.
At this level, syntax becomes infrastructure. An open question, a binary choice, a tentative inference and an instruction each establish a different distribution of agency. An editable visual map creates a different relationship to knowledge from a machine-generated conclusion. A request for confirmation preserves uncertainty. A pause may protect space for reflection. Every formal choice allocates agency, attention, authority and interpretive freedom.
This is not a speculative requirement; we keep colliding with it. In our own reflective systems we have had to specify when the same question is lawful and when it is not — the invitation that opens a morning’s thinking, and the identical sentence that, late at night, after a day of refusals, becomes pressure. What resolves the difference is not two prompts but one rule, and someone had to write it. We have had to build a first generative grammar for the prompts a machine may send a person, in which the warrant for speaking now must be argued inside the ask rather than assumed, every claim about a person must trace to something they actually said or did, and silence is the designed default. We have had to design how a shared field of commitments changes when one participant withdraws.
Each time, we found no profession to hire from. The work fell between interaction design, which owns the encounter; conversation design, which owns the script; and systems engineering, which owns the state — and belonged fully to none of them. We assembled the discipline ad hoc, and the assembly is the evidence: the capability gap is real, because we keep having to fill it by hand.
The capability we need therefore sits at the intersection of linguistics, communication, visual and interaction design, computational logic and contextual interpretation. It might be embodied by someone who has studied linguistics alongside visual communication: someone able to move between semantics, pragmatics, visual hierarchy, conversational rhythm and parametric system behaviour.
The essential capability is spacetime-contingent thinking: the ability to understand how meaning changes across time, place, relationship, culture, institutional setting and psychological disposition — and to translate those contingencies into governable, parametric form without ever mistaking the translation for the situation.
The meta-designer shapes how a system senses, hesitates, asks, adapts and learns. They determine how interactions condition one another, how an individual response alters a shared field, and how machine initiative should expand, contract or withdraw as circumstances change. They design the conditions under which the person, the conversation, the collective situation and the artefact can evolve together.
The deliverable of this practice is concrete: a written grammar — inspectable, contestable, revisable — from which a system’s possible moves derive and against which its actual moves can be judged. Because the deliverable is an artefact rather than an intuition, the capability can be held by a team as well as embodied in a person.
Meta-design is attentive design for changing situations. Its working principles are negotiated co-presence, participatory legitimacy and an evolving grammar of situated response. It allows multiple interactions to remain coherent while preserving difference, revision and refusal.
This is a capability gap already being exposed by our work.
If you have been developing this practice — as an individual whose work crosses linguistics, communication, visual design, computational systems and situated judgement; as a team that has had to invent it mid-project; or as a practice that has been doing it under another name — please get in touch with us. This may be a field we now need to build together.
A note on this pageIn the spirit of honesty: this piece would itself be much better with a meta-designer. It meets every reader in the same form, at the same register, at whatever hour they arrive—and it cannot hear what they would say back. It was designed. It was not yet meta-designed.