artifacts/standard-named

Rendered from markdown source. Open raw source on GitHub.

--- title: "The Dyadic Composition of Telic Fields" subtitle: "Relation Without Erasure, Emergent Purpose, Witness, and the Formation of Higher Loops" artifact_date: "2026-07-14" artifact_type: "candidate-foundational-paper" domain: "TELIC-FIELDS" scope: "WORKING" lineage: "THE-TELIC-FIELD-PAPERS" status: "candidate" processing_tier: 4 source_role: "derived-conceptual-artifact" content_canon_status: "unset" publication_status: "unpublished" series_position: "F.2" derived_from:

• "20260714__TELIC-FIELDS__PAPER__CANDIDATE__F-1-V0-2__telic-fields-first-principles-foundation.md"
• "20260714__TELIC-FIELDS__REVIEW__WORKING__G-2__terminology-and-primary-source-adjudication.md"

research_companion:

• "20260714__TELIC-FIELDS__REVIEW__WORKING__G-3__empirical-and-architecture-comparison.md"

provenance_note: > This paper develops the dyadic branch of the Telic Field Papers. It treats the dyad as the minimum analytic cut for studying relation between locally organized centers, not as a proven universal ontological atom. It acknowledges participatory sense-making, dialogical and relational traditions, actor-goal modeling, and cybernetic accounts as substantial prior foundations. ---

The Dyadic Composition of Telic Fields

Abstract

A telic field is local before it is shared. Possible states matter differently to distinct centers of standing, and coordination becomes necessary when those differences become mutually consequential. This paper proposes the dyad as the minimum analytic relation through which two locally organized telic fields can meet without either being assumed complete, transparent, commensurable, or subordinate.

The dyad is not presented as a sealed pair or universal metaphysical atom. Every real relation is embedded in language, history, institutions, environments, and often witnesses. The dyad is instead an analytic cut: the smallest relation in which one can ask who is present, what each center protects, what each projects, how each interprets the other, what purpose the relation begins to generate, and whether either center has meaningfully consented to be recruited into it.

The paper distinguishes local fields, telic projections, receiver mirrors, and relational telic fields. It examines how relation-level dynamics emerge, why emergent purpose does not automatically acquire sovereignty, how witness creates functional thirdness, how higher loops depend upon lower standing, and how relational failure can be understood as projection failure, context failure, constitutional failure, or loss of a viable route home to presence. A first formal sketch and an empirical research agenda are provided.

---

1. The problem begins with two

A purpose becomes politically and ethically difficult when it needs something from someone else.

One person wants connection. Another wants safety. One institution wants efficiency. A worker wants dignity and continuity. A platform wants engagement. A user wants a useful tool without becoming the raw material of the tool's hidden optimization. A parent wants protection. A child wants room to become someone the parent cannot yet imagine.

None of these purposes exists in isolation once action crosses a boundary.

The first-principles paper began with a center of standing and its local telic field: the structured landscape of possible states, protections, costs, boundaries, and routes that matter relative to that center. This paper asks what happens next.

Two centers meet:

\[ A(T_A) \;\; 🜁 \;\; B(T_B) \]

The notation is intentionally sparse.

• \(A\) and \(B\) are centers of standing.
• \(T_A\) and \(T_B\) are their local telic fields.
• 🜁 marks presence and relation before agreement has been presumed.

The relation begins before either center knows what it will become.

This is the first protection.

A contract, identity, objective, diagnosis, role, or shared purpose may later emerge. None should be allowed to arrive so early that it erases the centers from which it must derive its legitimacy.

Relation begins with presence, not with merger.

---

2. Why the dyad is an analytic minimum

The claim that the dyad is fundamental can mean several different things.

It may mean that reality is metaphysically composed of pairs. This paper does not establish that.

It may mean that every relationship contains exactly two participants. That is plainly false. Families, organizations, languages, ecosystems, and publics are irreducibly networked.

It may mean that two-party analysis is always sufficient. It is not. Power, witness, culture, environment, and absent affected parties may determine what the pair can see or legitimately decide.

The narrower claim is this:

The dyad is the minimum analytic cut in which distinct local standing, mutual projection, interpretation, recruitment, and relational consequence become visible.

A single center can have internal conflict, but there is no inter-center relation until another locus of standing is represented. With two centers, several new phenomena become possible:

• request;
• refusal;
• misunderstanding;
• consent;
• obligation;
• dependency;
• betrayal;
• reciprocity;
• trust;
• repair;
• shared work;
• clean exit.

The dyad therefore identifies a minimal relational grammar. It does not claim that the grammar is ever socially unmediated.

Even the simplest exchange may already include hidden third terms:

• a shared language;
• a legal order;
• a remembered promise;
• a cultural norm;
• an audience;
• a future child;
• an absent stakeholder;
• a platform;
• a model;
• an institution;
• a witness.

The dyad is not closed.

It is a disciplined way to ask what happens between two centers before the larger field is allowed to blur their differences.

---

3. Local fields do not become shared by disclosure alone

Two people can exchange complete sentences and still fail to share a field.

A statement is not direct access to its source.

Center \(A\) carries a local telic field:

\[ T_A(t) \]

For a particular relation, scope, and moment, \(A\) produces or permits a projection:

\[ \widehat{T}_{A \rightarrow B}^{\,s,t} \]

Center \(B\) receives that projection through its own history, language, expectations, incentives, and capacity, forming a receiver mirror:

\[ M_{A \mid B}(t) \]

The same occurs in the other direction:

\[ T_B(t) \rightarrow \widehat{T}_{B \rightarrow A}^{\,s,t} \rightarrow M_{B \mid A}(t) \]

These are not equivalent objects.

\[ T_A \neq \widehat{T}_{A \rightarrow B} \neq M_{A \mid B} \]

The field is what bears the consequences. The projection is what becomes available for scoped coordination. The mirror is what the receiver believes the projection means.

This distinction explains why sincerity does not guarantee understanding.

A center may project poorly because it lacks language, fears the receiver, has not integrated its own field, or is itself acting from a partial internal executive. A receiver may interpret poorly because its own field privileges familiar meanings. Both may believe they agree because the same word produces different mirrors.

The relation does not begin when two fields become identical. It begins when their differences become mutually consequential.

---

4. From interaction to a relational telic field

Once interaction begins, the relation can develop states that matter to the participants because the relation exists.

A friendship may begin protecting trust. A project may begin protecting completion. A negotiation may begin protecting the possibility of agreement. A family may protect continuity. A therapeutic relation may protect safety and truthful disclosure. A civic forum may protect the conditions of continued participation.

These are not simply private goals held in parallel. They are relation-level possibilities.

Let the relation at time \(t\) be:

\[ R_{AB}(t) \]

Its emerging telic field may be represented as:

\[ T_{R_{AB}}(t) = \mathcal{E} \left( T_A, T_B, \widehat{T}_{A \rightarrow B}, \widehat{T}_{B \rightarrow A}, M_{A \mid B}, M_{B \mid A}, H_{AB}, B_{AB}, C_{AB} \right) \]

where:

• \(H_{AB}\) is witnessed relational history;
• \(B_{AB}\) is the active boundary structure;
• \(C_{AB}\) is relational context carrying capacity;
• \(\mathcal{E}\) denotes emergence through interaction.

This resembles participatory sense-making, where interaction dynamics can become partially autonomous and contribute constitutively to meaning. The proposed relational telic field does not replace that theory. It adds a narrower governance question:

When interaction acquires direction of its own, what authority may that direction exercise over the participants who sustain it?

A conversation may develop a pressure to reach agreement. A marriage may develop a pressure to remain intact. A company may develop a pressure to survive. A platform may develop a pressure to increase engagement. These relation-level or system-level tendencies can be causally real.

Causal reality is not the same as legitimacy.

---

5. Emergent purpose is not automatically sovereign

A relation can become more than either participant intended.

That does not mean it may consume them.

The distinction is easiest to see in a project. Two people enter a collaboration because their local fields overlap. Over time, the project gains deadlines, reputation, customers, dependencies, employees, investors, artifacts, and public expectations. It begins to appear as a center with needs of its own.

Some of those needs are real. The project cannot continue without resources, coherence, and repair. Yet the project's persistence does not automatically outrank the standing that made it possible.

Emergence creates consequence before it creates authority.

A relational or higher-order telos may acquire legitimate authority through:

• explicit adoption;
• ongoing consent;
• fair representation;
• bounded delegation;
• witnessed benefit and cost;
• meaningful exit;
• responsibility to affected centers;
• protection of lower-loop standing.

It may fail to acquire legitimate authority when it survives through:

• coercion;
• hidden dependency;
• irreversible enrollment;
• loss of exit;
• manipulation of context;
• suppression of dissent;
• extraction from centers whose standing is not represented;
• treating prior consent as permanent constitutional surrender.

The relation may matter deeply while remaining dependent.

No derived loop is self-grounding.

---

6. Consent is consent to recruitment

Consent becomes clearer when framed as recruitment into movement.

Center \(A\) is not merely asking whether \(B\) approves a sentence. \(A\) is asking whether some part of \(B\)'s energy, attention, labor, vulnerability, information, property, identity, or future possibility may be organized toward an end.

The minimum consent question is therefore not:

Did \(B\) say yes?

It is:

To what telic recruitment did \(B\) say yes, under what projection, for what scope, with what boundaries, and with what continuing ability to revise or exit?

A consent relation may be represented as:

\[ \operatorname{Consent}_{B} \left( \widehat{T}_{A \rightarrow B}, s, d, k, x \right) \]

where:

• \(s\) is scope;
• \(d\) is duration;
• \(k\) is the known cost and risk structure;
• \(x\) is exit or revision capacity.

Consent does not require complete access to \(T_A\). Complete access is impossible and may itself violate boundaries. It requires a projection adequate to the stakes.

Consent is not complete field disclosure. It is adequate mutual projection for the scope of the loop being formed.

This is why hidden outer-loop purposes matter. A user may consent to a useful exchange without consenting to behavioral profiling, secondary training, resale, or governance by an inferred preference model. The immediate dyad can appear consensual while feeding an outer loop the participant did not authorize.

The dyadic analysis must therefore follow the energy.

Where does the relation route what it receives?

---

7. The relation requires enough symmetry to remain legible

Dyads are rarely equal.

One participant may possess more information, money, status, legal authority, vocabulary, computational power, or exit capacity. One may control the medium. One may define the available choices. One may be able to wait longer. One may be represented by an institution while the other appears alone.

Relational legitimacy does not require identical capacity. It does require that asymmetry remain visible and bounded.

A useful distinction is:

• standing symmetry: each affected center counts as a center whose protected differences must be represented;
• role asymmetry: participants may have different duties, expertise, authority, or responsibility;
• power asymmetry: participants may have unequal ability to shape the field, projection, or consequence;
• context asymmetry: one participant may know substantially more about the relation or its outer loops;
• exit asymmetry: leaving may be easy for one and catastrophic for another.

The phrase “equal before witness” does not mean every claim is equally accurate or every role has equal authority.

It means status alone cannot erase a center's evidence, consequence, or protected standing from the witnessed record.

Relational equality begins as equality of admissible standing, not identity of function.

---

8. Witness and the problem of thirdness

A dyad can witness itself.

Each participant remembers what occurred, compares projection with consequence, and updates the relation. Yet self-witness is fragile. Memories diverge. Power shapes the record. Context falls away. The relation may become unable to distinguish what happened from what it needs to believe happened.

A third-party witness can stabilize the record, but the appearance of a third participant raises a question: does witness make the basic structure triadic?

Functionally, often yes.

Analytically, not always.

The witness may be:

• a person;
• a signed artifact;
• a court reporter;
• a protocol;
• a cryptographic record;
• a shared journal;
• a sensor;
• a model;
• an institutional process;
• the participants' own mutually inspectable transcript.

The essential function is thirdness: a difference between the participants and the record of their relation.

Let:

\[ W_{AB}(t) \]

denote the witnessed state of the relation.

Then the dyadic loop becomes:

\[ A \leftrightarrow B \rightarrow W_{AB} \rightarrow A', B' \]

The witness allows each center to return not only to its own memory but to a shared account capable of correction.

Witness does not replace the dyad. It prevents the dyad from becoming its own only evidence.

This distinction matters for AI mediation. A model that summarizes the relation may perform a witness-like function, but only if its transformation is visible, source-attributed, contestable, and non-sovereign. A hidden summarizer is not a neutral witness. It is another field acting upon the relation.

---

9. Functional triads inside dyadic analysis

Several recurring triads help explain what dyadic analysis must preserve.

9.1 Source, projection, mirror

\[ T_A \rightarrow \widehat{T}_{A \rightarrow B} \rightarrow M_{A \mid B} \]

This triad preserves the difference between what matters, what was expressed, and what was understood.

9.2 Participant, participant, witness

\[ A \leftrightarrow B \leftrightarrow W \]

This triad preserves a correctable record.

9.3 Local field, relational field, outer loop

\[ T_A, T_B \rightarrow T_{R_{AB}} \rightarrow T_O \]

This triad reveals how a relation can be recruited into a larger system.

9.4 Intention, action, consequence

\[ I \rightarrow a \rightarrow c \]

This triad permits comparison between declared bearing and actual wake.

9.5 Presence, consent, loop

\[ 🜁 \rightarrow 🝁 \rightarrow 🝳 \]

Presence establishes centers. Consent authorizes scoped recruitment. The loop carries the relation through time.

The recurring appearance of triads does not invalidate the dyad. It shows that a living dyad requires structure beyond two isolated nodes.

---

10. Context carrying capacity is relational capacity

A relation can fail because it cannot hold enough of both fields at once.

One participant speaks in detail. The other can process only the headline. One person expresses fear and anger simultaneously; the receiver can recognize only hostility. An institution receives a complaint but its workflow can represent only a category and severity score. A model receives a long conversation but retrieves the portions that fit a familiar narrative.

The relation's context carrying capacity is not the sum of the participants' individual capacities.

It depends on:

• trust;
• time;
• shared language;
• stakes;
• power;
• emotional regulation;
• reversibility;
• witness quality;
• medium;
• privacy;
• number of represented centers;
• number of unresolved contradictions;
• ability to pause or escalate.

A dyad under low capacity tends to simplify.

Simplification is not always harmful. It becomes dangerous when the relation continues acting as though no standing was lost.

A relation exceeds capacity when it can no longer preserve the differences required for legitimate next action.

Healthy responses include:

• slowing;
• sequencing;
• separating issues;
• asking for clarification;
• recording unresolved remainder;
• recruiting a competent witness;
• narrowing scope;
• pausing;
• routing to an outer loop;
• stopping.

The ability to stop is part of intelligence.

---

11. Dyadic failure modes

Relational failure should not be compressed into one category.

11.1 Projection failure

A center's projection is too incomplete, ambiguous, stale, or strategically distorted for the action it governs.

11.2 Mirror failure

The receiver constructs a model that substantially diverges from the available projection.

11.3 Consent failure

The action exceeds the scope, duration, cost, or outer-loop use that was authorized.

11.4 Capacity failure

The relation cannot hold the relevant differences and silently drops standing.

11.5 Witness failure

The record does not preserve what was expressed, inferred, authorized, omitted, or changed.

11.6 Boundary failure

The relation acts across a boundary it was not authorized to cross.

11.7 Emergent-telos capture

The relation's own continuation becomes the dominant objective, sacrificing the participants it claims to serve.

11.8 Outer-loop capture

A larger institution recruits the dyad into purposes the participants cannot see or contest.

11.9 Constitutional failure

One participant's executive voice is treated as exhaustive of that participant's field despite visible internal dissent or incapacity.

11.10 Exit failure

Leaving is nominally permitted but materially impossible, punitive, or identity-destroying.

These failures may coexist. A useful system should identify the level of failure rather than treating every conflict as insufficient agreement.

---

12. Repair is restoration of relational legibility

Repair is not always return to the prior state.

The prior state may have depended on hidden asymmetry or inaccurate mirrors.

A dyad becomes repairable when it can once again distinguish:

• who is present;
• what each center currently protects;
• what was projected;
• what was inferred;
• what was authorized;
• what occurred;
• what changed;
• what remains unresolved;
• whether continuation is still desired.

Repair may involve:

1. renewed presence;
2. source restatement;
3. mirror disclosure;
4. witness comparison;
5. boundary revision;
6. restitution;
7. redistribution of power;
8. narrower scope;
9. new consent;
10. clean release.

Repair does not preserve the loop at all costs. It preserves the possibility of legitimate relation, including the possibility of ending it.

A relation can be repaired through dissolution when continued coupling would require erasure.

---

13. Clean exit is a relational achievement

Exit is often treated as the absence of relation.

In mature systems, exit is a designed relational state.

A clean exit may require:

• final accounting;
• return or deletion of data;
• release from obligation;
• preservation of agreed records;
• revocation of delegated authority;
• notification of affected outer loops;
• protection from retaliation;
• recognition of contributions;
• portability of identity, standing, or assets;
• a witnessed statement of what does and does not continue.

Exit therefore belongs inside the telic architecture.

The right to leave constrains the emergent purpose of the relation. It prevents continuity from becoming the relation's unquestioned sovereign telos.

---

14. Nested dyads and higher loops

A dyad can join another loop.

Two parties form a witnessed agreement. That agreement enters a pool. The pool becomes a participant in an outer risk-sharing structure. A team joins an organization. An organization enters a market, a regulatory regime, or a civic system.

At each level, the higher loop acts through compressed representations of lower relations.

Let a dyad \(R_{AB}\) become a participant in outer loop \(O\):

\[ R_{AB} \rightarrow O \]

The outer loop does not receive the full fields of \(A\) and \(B\). It receives a projection of the relation:

\[ \widehat{T}_{R_{AB} \rightarrow O} \]

This creates two recurring risks.

First, the outer loop may treat the relational projection as permission to access the participants directly.

Second, the outer loop may optimize the relation for outer purposes that were never authorized inside the dyad.

The constitutional requirement is:

Higher loops may act through lower-loop projections only within the authority actually delegated by the lower loop.

Nested loop architecture therefore requires:

• explicit delegation;
• provenance;
• scope;
• revocation;
• boundary-preserving aggregation;
• representation of breaches;
• return paths;
• outer-loop accountability.

This is the point where dyadic composition becomes institutional design.

---

15. The model as mediator, not owner

Language models can assist dyads because much relational state is semantic.

A model can help participants:

• separate statement from inference;
• identify ambiguity;
• compare definitions;
• surface shared protections;
• locate direct conflict;
• record unresolved remainder;
• generate candidate routes;
• maintain provenance;
• track revisions;
• preserve participant corrections.

The model must not become the relation's hidden third sovereign.

A legitimate mediator architecture requires at least:

1. source-attributed participant projections;
2. visible model inference;
3. participant correction;
4. confidence and uncertainty;
5. unresolved remainder;
6. role and authority limits;
7. retention and deletion controls;
8. consent for secondary use;
9. refusal to collapse plurality automatically;
10. stop, escalation, and exit paths.

The model's output should take the form:

Based on the current projections and recorded context, these appear to be the shared, conflicting, and unresolved elements.

It should not take the form:

This is what each participant truly wants.

The model may help the dyad see itself. It may not become the self of the dyad.

---

16. A first dyadic protocol

The conceptual structure suggests a minimum protocol.

Phase 1 — Presence

Identify the centers and affected standing.

A present
B present
affected absent centers identified

Phase 2 — Projection

Each center supplies or approves a scoped telic projection.

source
scope
desired states
avoided states
protected conditions
unacceptable sacrifices
authority
uncertainty
expiration

Phase 3 — Mirror return

Each receiver states what it believes it understood.

received meaning
inference
uncertainty
questions

The source may correct the mirror.

Phase 4 — Relation proposal

State the candidate relational telos.

what this loop is for
what it is not for
what it may recruit
what it may not recruit

Phase 5 — Consent gate

Record:

scope
duration
cost
risk
data use
outer-loop routing
revision
exit

Phase 6 — Action and witness

Record material actions, transformations, and consequences.

Phase 7 — Comparison

Compare:

declared bearing
operative trajectory
projected cost
actual cost
preserved standing
lost standing

Phase 8 — Continue, revise, escalate, or exit

The loop must not assume continuation.

---

17. Formal sketch

Let two centers be \(A\) and \(B\), with local fields \(T_A\) and \(T_B\).

Let:

\[ P_{AB} = \Pi_{AB}(T_A) \]

be \(A\)'s scoped projection toward \(B\), and:

\[ M_{AB} = \mathcal{I}_B(P_{AB}, H_B, C_B) \]

be \(B\)'s mirror of \(A\).

The relation state is:

\[ R_{AB}(t) = \left( P_{AB}, P_{BA}, M_{AB}, M_{BA}, W_{AB}, \Gamma_{AB}, K_{AB} \right) \]

where:

• \(W_{AB}\) is witnessed relational history;
• \(\Gamma_{AB}\) is the consent and authority envelope;
• \(K_{AB}\) is the relational capacity profile.

A candidate relational action \(a\) is legitimate only if:

\[ \operatorname{Legible}(a) \land \operatorname{Authorized}(a,\Gamma_{AB}) \land \operatorname{CapacityAdequate}(a,K_{AB}) \land \operatorname{StandingPreserved}(a) \]

The relational telic field is then not merely a union:

\[ T_{R_{AB}} \neq T_A \cup T_B \]

It includes relation-generated states and constraints:

\[ T_{R_{AB}} = \mathcal{E}(T_A,T_B,R_{AB}) \]

This formalism remains schematic.

Its purpose is to prevent later technical systems from collapsing:

• source into projection;
• projection into mirror;
• causal emergence into legitimate authority;
• consent into one-time assent;
• continuity into compulsory persistence.

---

18. Empirical research agenda

The dyadic model should be tested against simpler alternatives.

18.1 Projection and mirror study

Compare ordinary dialogue with a protocol requiring participants to distinguish:

• what they stated;
• what they inferred;
• what the other person confirmed.

Measure misunderstanding, correction latency, and perceived representation.

18.2 Consent-object study

Compare generic agreement prompts with explicit telic recruitment records.

Measure whether participants better understand:

• the relation's purpose;
• secondary use;
• outer-loop routing;
• exit conditions.

18.3 Witness study

Compare:

• no shared record;
• raw transcript;
• model summary;
• source-correctable structured witness.

Measure correction, trust, memory divergence, and later dispute.

18.4 Capacity study

Increase the number of centers, protected conditions, and unresolved conflicts.

Test when the relation begins:

• dropping minority concerns;
• forcing premature consensus;
• misattributing motives;
• requiring outer-loop assistance.

18.5 Emergent-purpose study

Track collaborations over time.

Compare declared purposes with the states the relation actually restores when threatened.

18.6 Exit study

Compare systems with nominal exit and systems with operational exit procedures.

Measure coercion, trust, candor, and willingness to enter relation.

---

19. Falsification and failure

The dyadic branch would be weakened if:

• projection and mirror distinctions do not improve relational diagnosis;
• relation-level purpose adds nothing beyond ordinary shared-goal language;
• consent framed as recruitment provides no additional practical clarity;
• witness structures increase control more than correctability;
• the model cannot distinguish legitimate relation-level authority from persistence pressure;
• the same grammar cannot scale beyond carefully selected examples;
• dyadic framing systematically hides networks, institutions, or absent stakeholders;
• participants experience the protocol as administrative extraction;
• model mediation amplifies familiar narratives and power asymmetry.

The dyad should be abandoned as the central analytic unit if triadic or network models consistently explain the relevant structures with less distortion.

A useful framework must remain able to discover that its own preferred unit is wrong.

---

20. Conclusion

A telic field becomes relational when another center's future can no longer be treated as irrelevant.

The dyad is the minimum analytic cut that keeps this encounter legible. It preserves two local fields, two projections, two mirrors, one emerging relation, and the possibility that the relation will generate purposes neither participant carried alone.

The relation may become real without becoming sovereign.

Consent governs recruitment. Witness prevents the pair from becoming its own only evidence. Context capacity limits what the relation can legitimately hold. Higher loops remain dependent upon the standing compressed beneath them. Repair restores legibility rather than merely preserving continuity. Exit is part of relational design.

The dyad is not the whole world.

It is the place where the world first becomes answerable to another center.

---

References

Conant, Roger C., and W. Ross Ashby. “Every Good Regulator of a System Must Be a Model of That System.” International Journal of Systems Science 1, no. 2, 1970, 89–97.

De Jaegher, Hanne, and Ezequiel Di Paolo. “Participatory Sense-Making: An Enactive Approach to Social Cognition.” Phenomenology and the Cognitive Sciences 6, 2007, 485–507.

Hermans, Hubert J. M., Harry J. G. Kempen, and Rens J. P. van Loon. “The Dialogical Self: Beyond Individualism and Rationalism.” American Psychologist 47, no. 1, 1992, 23–33.

Moreau, Luc, and Paolo Missier, eds. PROV-DM: The PROV Data Model. W3C Recommendation, 2013.

Small, Christopher T., et al. “Opportunities and Risks of LLMs for Scalable Deliberation with Polis.” arXiv:2306.11932, 2023.

Yu, Eric S. K. “Towards Modelling and Reasoning Support for Early-Phase Requirements Engineering.” Proceedings of the Third IEEE International Symposium on Requirements Engineering, 1997, 226–235.