# Research Architecture: Intelligent Voluntary Cooperation

## Evidence system for the Qi + IXO white paper

**Status:** Working research specification  
**Companion:** [Research Evidence Dossier](./research-evidence-dossier.md)  
**Scope:** The theory, evidence, system mapping, and empirical program behind the proposed white paper *Intelligent Voluntary Cooperation: How shared intelligent environments enable humans, agents and things to turn intent into reality*.

---

## 1. Purpose

The narrative outline makes a large argument: persistent shared environments may coordinate autonomous humans and agents without requiring every useful action to be centrally assigned, and Qi + IXO may provide the cooperative and consequential substrate for this mode of coordination.

The research architecture exists to keep that argument rigorous. It prevents five different kinds of statement from being blended together:

1. what prior research has established;
2. what a particular experiment observed;
3. what follows as a defensible theoretical inference;
4. what Qi and IXO demonstrably implement today; and
5. what remains a product, strategic, or scientific hypothesis.

The intended outcome is not a promotional paper with academic decoration. It is a testable theory of intelligent voluntary cooperation, a transparent mapping from that theory to a computational architecture, and a research program capable of showing where the architecture works, where orchestration works better, and where the thesis fails.

---

## 2. The target contribution

### Central proposition

> Under appropriate boundaries, autonomous participants can coordinate useful action by perceiving and modifying a persistent shared environment. For such coordination to become voluntary, trustworthy, and consequential in human settings, the environment must also make intent, capability, authority, provenance, evaluation, and legitimate consequence legible.

### The Qi + IXO proposition

> Qi is a candidate cooperative environment in which humans and intelligent agents encounter shared intent, state, artifacts, and affordances. IXO is a candidate consequential substrate through which identity, authority, claims, evidence, decisions, and state transitions can be made trustworthy.

The phrase **candidate** is deliberate. The white paper must demonstrate current implementation claims from system evidence and present unimplemented or unvalidated claims as design hypotheses.

### Contribution stack

| Layer | Contribution | Required support |
|---|---|---|
| Empirical | Persistent shared environments can mediate decentralized coordination and artifact inheritance under specified conditions. | Primary experiments and their limitations |
| Theoretical | Stigmergy, distributed cognition, affordances, institutional governance, and agentic systems can be combined into a theory of intelligent voluntary cooperation. | Explicit synthesis; no false equivalence between literatures |
| Computational | The theory can be expressed as actors operating on bounded state through legitimate actions that create traces, evaluations, decisions, and consequences. | Definitions, state model, causal mechanism, failure modes |
| Architectural | Qi + IXO can be evaluated as an implementation of that model. | Repositories, protocols, schemas, running behavior, and system owners |
| Strategic | Accumulated cooperative artifacts and consequential histories may create compounding utility and defensibility. | Clearly marked hypothesis plus longitudinal evidence plan |

---

## 3. The causal model

The paper should be built around a mechanism, not a metaphor.

```text
bounded intent and context
          ↓
participants perceive state and affordances
          ↓
participants choose legitimate actions
          ↓
actions create or modify persistent traces
          ↓
traces change what later participants can perceive and do
          ↓
claims and evidence pass through evaluation and authority
          ↓
accepted outcomes produce consequential state
          ↓
new state exposes new affordances and reusable structures
          ↺
```

The mechanism only counts as intelligent voluntary cooperation when all of the following are present:

- **agency:** participants can act, not merely receive outputs;
- **legibility:** intent, relevant state, and possible contributions can be perceived;
- **choice:** participation and contribution are not exhausted by central assignment;
- **bounded authority:** actors can only produce effects within legitimate capabilities;
- **persistence:** useful traces can outlive the action or actor that produced them;
- **evaluation:** assertions are distinguishable from accepted outcomes;
- **consequence:** accepted decisions can change a shared reality;
- **governance:** participants can contest, correct, retire, or constrain harmful state;
- **exit:** voluntary participation includes a meaningful ability to decline or leave, subject to legitimate prior commitments.

### Boundary conditions

The theory does **not** predict that environmental coordination is universally superior. Orchestration is expected to remain preferable when execution is tightly coupled, the correct sequence is known, timing is critical, or a single accountable authority must control the result.

The research question is therefore comparative:

> Which coordination mode, or hybrid of modes, produces the best outcomes under which task, risk, and institutional conditions?

---

## 4. Working vocabulary

These terms must be fixed before drafting because imprecision here would infect the whole paper.

| Term | Working definition | Important exclusion |
|---|---|---|
| Actor | A human, organisation, software agent, device, or governed service that can perceive state and take actions. | A model alone is not necessarily an actor. |
| Environment | The bounded medium containing observable state, rules, resources, and action possibilities. | Not merely a chat history or database. |
| Intent | A declared future condition toward which cooperation is oriented. | Not proof that participants share motives. |
| State | The currently represented condition of the environment. | Representation is not automatically truth. |
| Trace | A state change left by an action that can influence later action. | Not every stored event is relevant or legible. |
| Artifact | A persistent, addressable trace with informational, executable, institutional, or consequential function. | Persistence alone does not make an artifact useful. |
| Affordance | An action possibility available to a particular actor given the state, its capabilities, and its authority. | Not a generic task visible equally to everyone. |
| Capability | What an actor is able to do. | Ability does not imply permission. |
| Authority | What an actor is permitted to decide or effect within a scope. | Identity does not imply authority. |
| Claim | An assertion about an event, condition, contribution, or entitlement that may require evaluation. | A claim is not an accepted fact. |
| Evidence | Material used under a stated method to support, weaken, or refute a claim. | Data does not interpret itself. |
| Decision | A determination made under an identifiable rule, method, or authority. | Model output is not automatically a decision. |
| Consequence | An effective change in rights, resources, credentials, permissions, obligations, or authoritative state. | A recommendation is not a consequence. |
| Cooperation | Actors combining or aligning contributions toward a shared or compatible outcome. | Mere co-presence or parallel activity is insufficient. |
| Voluntary | Actors can make informed contribution choices within legitimate boundaries and retain meaningful refusal or exit. | “Not centrally assigned” is insufficient. |
| Orchestration | Coordination through explicit decomposition, assignment, sequencing, and aggregation by a controlling process or actor. | It can be distributed technically while remaining centrally specified. |
| Stigmergic coordination | Indirect coordination in which traces of action in a medium influence later action. | It does not explain all human cognition, institutions, or motives. |
| Reality boundary | The procedures and authorities through which assertions become accepted consequential state. | It does not imply metaphysical or universal truth. |

### Terms that must remain distinct

- collaboration, coordination, cooperation;
- agent, model, actor, participant;
- artifact, message, event, state;
- capability, authority, permission;
- assertion, claim, evidence, decision, fact;
- persistence, authoritativeness, consequence;
- self-organization, voluntariness, legitimacy;
- intelligence, agency, autonomy.

---

## 5. Primary research questions

### RQ1 — Coordination mechanism

Under what conditions can persistent environmental state substitute for, reduce, or complement direct communication and central orchestration?

### RQ2 — Artifact inheritance

When do persistent informational or executable artifacts improve later participants' performance, especially across participant turnover?

### RQ3 — Voluntary contribution

What must an environment expose for heterogeneous actors to discover, select, and legitimately perform useful contributions without central assignment?

### RQ4 — Bounded worlds

Can intent, scope, membership, state, rules, and completion conditions form a meaningful computational boundary for cooperation?

### RQ5 — Governance and safety

Which institutional and technical controls prevent poisoned state, capture, free riding, unauthorized consequences, and obsolete artifacts from degrading cooperation?

### RQ6 — Cognition and consequence

What mechanisms should separate what an agent proposes from what a system accepts as true or allows to become consequential?

### RQ7 — Hybrid coordination

Which tasks benefit from stigmergic coordination, which require orchestration, and how should systems move between them?

### RQ8 — Qi + IXO correspondence

Which parts of the proposed model exist in Qi and IXO today, which are partially implemented, and which remain conceptual?

### RQ9 — Cumulative cooperation

Does the accumulation of evaluated artifacts, protocols, identities, and contribution histories reduce the cost or increase the quality of future cooperation?

### RQ10 — Limits and falsification

At what scales, risk levels, or rates of state change do environmental coordination costs, information pollution, governance overhead, or power asymmetries overwhelm its benefits?

---

## 6. Claim and evidence discipline

Every substantive statement in the paper should carry one of these internal labels during drafting. The labels can be removed from the final prose only after the claim ledger is complete.

| Label | Meaning | Example form | Publication rule |
|---|---|---|---|
| **[E] Established evidence** | Repeated or canonical finding in an applicable literature. | “Persistent artifacts can mediate coordination in some human work settings.” | Cite primary research; state domain limits. |
| **[X] Experimental result** | Observation from a named experiment. | “In the reported SwarmWorld conditions, agents differentiated into behaviors.” | State setup, comparator, sample, metric, and limitation. |
| **[S] System fact** | Verifiable property of Qi or IXO today. | “The current Topic schema records X.” | Cite versioned source, protocol, schema, or reproducible behavior. |
| **[I] Inference** | The authors' synthesis or architectural interpretation. | “A Topic can be interpreted as a bounded cooperative environment.” | Show premises; avoid presenting it as external validation. |
| **[H] Hypothesis** | A proposition requiring empirical validation. | “Accumulated evaluated artifacts will reduce future coordination cost.” | Include test and falsifier. |
| **[N] Normative principle** | A design or governance commitment. | “Consequential actions should require bounded authority.” | Explain values and trade-offs; do not call it a scientific result. |

### Minimum claim record

Each important claim must have:

- a stable claim ID;
- exact wording narrow enough to verify;
- claim class;
- chapter and argument role;
- supporting sources or system evidence;
- contrary or limiting evidence;
- confidence level;
- scope conditions;
- consequence if wrong;
- status: proposed, sourced, challenged, revised, or retired.

### Confidence scale

| Level | Meaning |
|---|---|
| High | Directly supported by multiple applicable primary sources or reproducible system evidence. |
| Moderate | Supported, but transfer to this context requires an explicit inference. |
| Preliminary | Supported by one recent result, narrow evidence, or incomplete system archaeology. |
| Speculative | Plausible and useful to test, but not evidentially established. |

### Citation rules

1. Cite the source that owns the result: paper, specification, protocol, repository, or official dataset.
2. Do not use a press article, LinkedIn post, or vendor summary as evidence when the primary source exists.
3. Cite at claim granularity, not only at the end of a section.
4. Record study context, comparator, metric, and limitation for quantitative claims.
5. Treat preprints as preprints and recent single studies as preliminary.
6. Use canonical sources to define a field, but current systematic evidence to describe the strength and limits of a result.
7. For Qi + IXO, distinguish documented intent, implemented behavior, deployed behavior, and measured outcome.
8. A citation to adjacent research cannot validate a Qi-specific product claim.

---

## 7. Source architecture

### Tier 1 — Direct primary evidence

- peer-reviewed experiments and the exact SwarmWorld preprint, appendices, code, and data;
- original studies of human artifact-mediated coordination;
- Ostrom's own institutional analysis and primary empirical work;
- official technical standards for identity, credentials, capabilities, CRDTs, and relevant protocols;
- versioned Qi and IXO source code, schemas, ADRs, protocol specifications, and reproducible product behavior;
- experiment logs and datasets produced by the Qi research program.

### Tier 2 — High-quality synthesis

- systematic reviews and meta-analyses;
- scholarly monographs and field-defining reviews;
- standards implementation guidance from authoritative bodies.

Use these to establish consensus, disagreements, and terminology—not to replace direct inspection of the primary result being relied upon.

### Tier 3 — Context and discovery

- research lab summaries;
- conference talks;
- credible technical essays;
- product documentation and public architecture explanations.

Useful for finding sources and understanding interpretation. They cannot independently carry a central claim.

### Tier 4 — Narrative context only

- social media posts;
- general media;
- uncited marketing pages;
- investor or community commentary.

These may establish why the question matters or how an idea is being discussed, but not whether the claim is true.

---

## 8. Research workstreams

### WS1 — Intellectual lineage and definition

**Question:** What is the minimum established theory needed to define the mechanism without turning stigmergy into a universal explanation?

**Evidence domains:** stigmergy, distributed cognition, situated action, affordance theory, CSCW, coordination mechanisms, collective intelligence, and cumulative culture.

**Outputs:**

- controlled vocabulary and conceptual boundary memo;
- literature map showing agreements and non-equivalences;
- evidence for and against transferring stigmergy from insects and software agents to human institutions;
- canonical examples of informational, executable, institutional, and consequential artifacts.

### WS2 — SwarmWorld verification

**Question:** What did the experiment actually establish?

**Required extraction:**

- environment, action schema, model, prompts, agent count, run count, seeds, baselines, controls, evaluation rules, disturbances, and outcome metrics;
- precise evidence for specialization, persistence, reuse, inheritance, resilience, and communication patterns;
- ablations and sensitivity analyses;
- the strongest null or negative result;
- what was simulator-determined versus model-judged;
- reproducibility status, code/data availability, and preprint status;
- explicit limits on generalizing to humans, open economies, real authority, or Qi.

**Output:** a claim-by-claim replication card, not a prose summary.

### WS3 — Institutional conditions for “voluntary”

**Question:** What makes decentralized participation legitimate and durable rather than merely unassigned?

**Evidence domains:** commons governance, polycentricity, collective-choice arrangements, membership and resource boundaries, monitoring, sanctions, dispute resolution, nested institutions, exit, power, and incentive compatibility.

**Outputs:**

- mapping from institutional conditions to computable mechanisms;
- analysis of where the mapping is faithful, partial, or misleading;
- explicit treatment of coercion, dependency, unequal bargaining power, and hidden orchestration;
- governance threat model.

### WS4 — Qi + IXO system archaeology

**Question:** What does the current system actually implement?

**Evidence order:**

1. protocol specifications and schemas;
2. architectural decision records and authoritative docs;
3. current source code and tests;
4. deployed, reproducible behavior;
5. product plans and design proposals, clearly marked as future state.

**Objects to trace:** Topic, Topic Contract, Matrix room/state/event, collaborative page/CRDT, entity identity, DID, capability or delegation, agent, request, Deed, Flow, POD, claim, evidence, evaluation, decision, credential, asset, payment, provenance, and chain state transition.

**For each object record:**

- canonical name and current definition;
- source of truth;
- lifecycle and state transitions;
- actor and authority model;
- persistence layer;
- links to other objects;
- user-visible behavior;
- implemented, partial, planned, or deprecated status;
- known gap against the theoretical model.

**Output:** an implementation correspondence matrix. No architectural claim enters the white paper without a row in this matrix.

### WS5 — Comparative coordination architectures

**Question:** What is gained and lost relative to alternatives?

Compare at least:

- single-agent execution;
- hierarchical multi-agent orchestration;
- peer-to-peer message coordination;
- shared-state environmental coordination;
- governed shared-state coordination;
- human-led and hybrid configurations.

Compare across task decomposability, need for global context, communication cost, latency, reliability, innovation, adaptability, accountability, security, participant turnover, and auditability.

**Output:** a conditional decision framework, not a universal ranking.

### WS6 — Trust, security, and failure modes

**Question:** What new harms arise when artifacts persist and influence later actors?

At minimum, investigate:

- state poisoning and prompt injection through artifacts;
- malicious or low-quality stigmergic traces;
- Sybil participation and reputation gaming;
- authority escalation and confused-deputy failures;
- cascading errors and feedback loops;
- stale-state anchoring and information pollution;
- capture, exclusion, collusion, and unequal visibility;
- unverifiable claims and evaluator compromise;
- irreversible consequences based on reversible cognition;
- privacy leakage and unwanted persistence;
- disputes over deletion, correction, and historical integrity.

**Output:** threat model linked to concrete architectural controls and residual risks.

### WS7 — Empirical Qi research program

**Question:** Which central claims can be tested in controlled and field settings?

**Outputs:** experimental protocols, preregistered hypotheses, instrumentation, benchmark tasks, datasets, ethics review, and publication criteria.

### WS8 — Adversarial synthesis

**Question:** What is the strongest case that the theory is unnecessary, incomplete, or wrong?

Required rival explanations include:

- apparent emergence is prompt-induced role play;
- environmental coordination merely hides orchestration in schemas and simulator rules;
- direct communication or simple shared memory explains the gains;
- persistent artifacts increase imitation while reducing independent discovery;
- governance costs erase coordination benefits;
- “voluntary” obscures incentives or structural coercion;
- blockchain or cryptographic infrastructure is unnecessary for most cooperation;
- Qi's concepts are a post-hoc mapping rather than a causal implementation.

**Output:** red-team memorandum and revised claims.

---

## 9. Core claim ledger

This is the minimum claim set the paper must source or test.

| ID | Claim | Class now | Evidence needed | Falsifier or limiting result | Main chapters |
|---|---|---|---|---|---|
| C01 | Coordination becomes a limiting problem as intelligent agency becomes cheaper and more abundant. | [I]/[H] | Measures of agent proliferation and coordination overhead; theory of coordination costs | No meaningful increase in coordination cost, failure, or governance burden | 1 |
| C02 | Central orchestration has structural limits and remains superior in defined conditions. | [E]/[I] | Distributed systems and multi-agent comparisons | Orchestration dominates across exploratory, adaptive, and turnover-heavy tasks | 2, 19 |
| C03 | Traces in a shared environment can mediate useful indirect coordination. | [E] | Canonical stigmergy and human/technical studies | Effects disappear outside narrow insect or toy settings | 3–4 |
| C04 | In SwarmWorld, persistent artifacts supported specialization, reuse, and technological inheritance under the reported conditions. | [X] | Full paper, appendix, code, data, baseline review | Result not reproducible or attributable to confounds | 5 |
| C05 | Persistent artifacts are a distinct and under-modeled unit of collective intelligence. | [I] | Cross-field synthesis and competing accounts | “Artifact” adds no explanatory or predictive value beyond shared memory/messages | 6 |
| C06 | Voluntary cooperation requires identity, informed choice, capability, bounded authority, incentives, and exit. | [I]/[N] | Institutional and political theory; empirical governance studies | Conditions are neither necessary nor predictive in target contexts | 7 |
| C07 | Actor-specific affordances can make useful contributions discoverable without central assignment. | [I]/[H] | Affordance theory plus controlled system experiments | No increase in contribution discovery, or hidden assignment explains outcomes | 8 |
| C08 | A bounded cooperative world needs the proposed minimum primitives. | [I]/[H] | Formal model, comparison, prototype ablations | Equivalent outcomes with a materially smaller or different primitive set | 9 |
| C09 | Persistence increases both coordination power and governance risk. | [E]/[I] | Security, memory, governance, and adversarial evidence | Risks are no greater than in ephemeral/message-based systems | 10–11 |
| C10 | Claims, evidence, evaluation, decision, and authority can separate cognition from consequence. | [I]/[S] | Formal semantics and verified IXO implementation evidence | Boundary can be bypassed, is circular, or adds no reliability | 12–13, 18 |
| C11 | Qi Topics can function as bounded environments for cooperation around intent. | [I]/[S]/[H] | Current schema/behavior plus user and agent studies | Topics remain passive containers and do not improve coordination | 14–15 |
| C12 | Deeds, Flows, and PODs can support executable inheritance. | [I]/[S]/[H] | Current implementation and longitudinal reuse evidence | Reuse is copy-only, non-composable, or yields no cumulative performance gain | 16 |
| C13 | Different state classes require different integrity and consensus mechanisms. | [I]/[N] | Threat/risk model and systems comparison | One mechanism is consistently better without material trade-offs | 17 |
| C14 | Evaluated contribution histories can improve future decisions, reputation, and cooperation. | [H] | Longitudinal field data | Histories create lock-in, gaming, bias, or no predictive lift | 18–20 |
| C15 | Accumulated cooperative state can create compounding utility and defensibility. | [H] | Cohort, reuse, switching-cost, and counterfactual data | Marginal utility does not rise; portability erases defensibility; pollution dominates | 20 |

---

## 10. Chapter-to-evidence map

| Chapter | Research burden | Evidence package | Required intellectual honesty |
|---|---|---|---|
| 1. Intelligence is becoming abundant. Coordination is not. | Establish the problem without assuming the conclusion. | Agent capability/cost trends; coordination failure and overhead literature. | “Coordination is the next bottleneck” is a thesis, not a settled fact. |
| 2. The orchestration paradigm | Describe dominant patterns and their trade-offs. | Multi-agent taxonomies, distributed coordination research, current agent frameworks. | Include cases where orchestration is the correct choice. |
| 3. The coordination paradox | Define stigmergy precisely. | Original/canonical definitions and modern clarification. | Do not equate all shared-state systems with stigmergy. |
| 4. From ant trails to human institutions | Establish transfer and its limits. | Human workplace, CSCW, distributed cognition, coordination-mechanism studies. | Human cooperation includes meaning, authority, norms, incentives, and communication. |
| 5. SwarmWorld | Report a recent experimental result. | Paper, appendix, code/data, author materials, independent reproduction when available. | One preprint supports a mechanism under test conditions; it does not validate Qi. |
| 6. The artifact | Build the artifact taxonomy. | Distributed cognition, software artifacts, cumulative culture, institutional records. | Show what “artifact” explains beyond storage and memory. |
| 7. Voluntary | Define legitimate choice. | Institutional economics, collective action, autonomy, consent, exit, power. | Self-organization is not automatically voluntary or legitimate. |
| 8. Shared affordances | Explain local discovery of possible action. | Affordance/situated action literature and prototype evidence. | Affordances are actor-relative and can be manipulated. |
| 9. Minimum primitives | Derive the computational ontology. | Formal definitions, comparison against alternatives, dependency analysis. | Present the set as a candidate until ablations support necessity. |
| 10. Why stigmergy is insufficient | Establish failure modes. | Security, adversarial multi-agent, persistent memory, information quality research. | Treat safety as part of the theory, not an implementation appendix. |
| 11. Ostrom meets agentic AI | Translate governance insights. | Ostrom primary work and later tests/refinements. | A mapping is not proof; common-pool resources are not identical to digital Topics. |
| 12. Agents may propose; reality must decide | Define the reality boundary. | SwarmWorld evaluator design, formal verification/evaluation, decision theory. | Deterministic evaluation is not available for every social claim. |
| 13. Agency + Decisions + Reality | Present the abstract architecture. | Formal state machine and authority model. | Distinguish descriptive model from branded system. |
| 14. Qi beyond chat | Establish current product trajectory. | Versioned architecture, schemas, code, and observed behavior. | Do not infer implementation from product language. |
| 15. Topics as bounded worlds | Test the central product correspondence. | Topic Protocol/contract, lifecycle, permissions, UX, agent behavior, user studies. | Define what makes a Topic more than a labelled room or thread. |
| 16. Deeds, Flows, PODs | Establish executable inheritance. | Object definitions, versioning, execution/evaluation/reuse traces. | Reusability and inheritance require observed lineage, not conceptual composability. |
| 17. Consequential substrate | Justify graduated integrity. | Threat models, state taxonomy, consistency/consensus trade-offs, IXO evidence. | Do not imply all persistence requires a blockchain. |
| 18. Claims as reality boundary | Demonstrate claim-to-consequence lifecycle. | Schemas, evaluation methods, authority checks, chain transitions, audit traces. | Accepted state is method- and authority-relative, not universal truth. |
| 19. Workflows to ecosystems | Compare known-path and emergent execution. | Comparative experiments and field cases. | Ecosystems can be wasteful, unequal, and hard to govern. |
| 20. Network effects | State and test the strategic hypothesis. | Reuse, retention, marginal utility, contribution, portability, and switching data. | “The moat is the world” remains speculative until longitudinally measured. |
| 21. Research agenda | Make the theory falsifiable. | Preregistered questions, metrics, datasets, and release plan. | Publish negative and boundary results. |

---

## 11. The argument bridge

The paper must cross four bridges explicitly. Most credibility risk sits here.

### Bridge A — Biology and social theory to human-agent systems

**Permitted inference:** environmental traces are a recurring coordination mechanism across several domains.

**Prohibited leap:** because ants coordinate stigmergically, human-agent institutions will self-organize effectively.

**Required evidence:** human artifact-mediated coordination, cognitive and institutional differences, and cases where the analogy fails.

### Bridge B — SwarmWorld to the general computational mechanism

**Permitted inference:** a controlled LLM-agent environment can test whether persistent shared artifacts support indirect coordination, specialization, and inheritance.

**Prohibited leap:** one simulated result proves persistent worlds are the general solution to multi-agent coordination.

**Required evidence:** exact experimental scope, baselines, ablations, replications, negative results, and rival explanations.

### Bridge C — General mechanism to Qi + IXO architecture

**Permitted inference:** the mechanism implies requirements such as bounded context, observable state, actor-relative affordances, persistent artifacts, authority, evaluation, and consequence.

**Prohibited leap:** similarly named Qi objects necessarily satisfy those requirements.

**Required evidence:** a traceable correspondence from every requirement to implemented schemas, rights, lifecycle, persistence, and user/agent behavior.

### Bridge D — Architecture to strategic value

**Permitted inference:** reusable, evaluated structures could reduce the cost of later cooperation.

**Prohibited leap:** accumulated state automatically creates network effects, defensibility, or a moat.

**Required evidence:** longitudinal reuse, performance lift, cross-context transfer, retention, portability, governance cost, and information-pollution measures.

---

## 12. Comparative empirical program

### Experimental conditions

Use matched tasks and resources across at least four conditions:

| Condition | Coordination available |
|---|---|
| O — Orchestrated | Central planner, explicit decomposition and assignment |
| M — Messaging | Peer communication and transient task context |
| S — Shared state | Persistent artifacts and local environmental observation |
| G — Governed shared state | Persistent artifacts plus identity, scoped capabilities, evaluation, dispute/correction, and consequence controls |

A fifth **hybrid** condition should allow orchestrated subflows inside a governed shared environment.

### Task families

- decomposable production with a known sequence;
- open-ended exploration and innovation;
- maintenance of a changing shared system;
- evidence collection and claim evaluation;
- resource allocation under conflicting incentives;
- long-horizon work with participant turnover;
- adversarial tasks containing poisoned or obsolete artifacts;
- mixed human-agent work where expertise and authority differ.

### Primary outcome measures

| Dimension | Candidate measures |
|---|---|
| Effectiveness | Completion rate, verified quality, utility, error rate |
| Coordination cost | Messages, tokens, planner calls, handoffs, duplicated work, time-to-contribution |
| Discovery | Useful opportunities identified; participation-to-contribution conversion |
| Adaptability | Recovery after disturbance; response to changed goals or resources |
| Inheritance | Reuse rate, lineage depth, modification quality, performance lift across generations |
| Diversity | Portfolio breadth and independence of solutions |
| Robustness | Tail performance, failure containment, participant-turnover sensitivity |
| Governance | Unauthorized actions, disputes, correction latency, capture and gaming indicators |
| Trust | Provenance coverage, evidence sufficiency, audit reconstruction, evaluator agreement |
| Voluntariness | Decline/exit rate, self-selection quality, informed-choice measures, coercion/dependency indicators |
| Economics | Cost per verified outcome, resource efficiency, value distributed per contribution |

### Initial preregisterable hypotheses

- **H1:** Shared-state conditions use less direct communication than messaging-only coordination for equally verified outcomes.
- **H2:** Shared-state and hybrid conditions outperform orchestration on adaptive, exploration-heavy, and participant-turnover tasks, but not on tightly sequenced deterministic tasks.
- **H3:** Persistent executable artifacts improve later-cohort performance relative to access to summaries alone.
- **H4:** Selective, governed persistence outperforms indiscriminate persistence by reducing stale-state and poisoning effects.
- **H5:** Actor-specific affordance presentation improves useful self-selection over a generic task list.
- **H6:** Evaluation and capability boundaries reduce unauthorized or unsupported consequential actions without eliminating most legitimate contributions.
- **H7:** The marginal value of accumulated artifacts is positive only when quality, retirement, provenance, and discovery mechanisms are effective.

### Key ablations

Remove or vary one element at a time:

- artifact persistence;
- direct communication;
- visible provenance;
- actor identity;
- scoped authority;
- evaluation before consequence;
- artifact search and ranking;
- correction and retirement;
- incentives;
- participant memory;
- explicit cultural transmission;
- central planner access.

### Field-pilot sequence

1. **Synthetic benchmark:** controlled agents, deterministic verification, repeatable seeds.
2. **Sandboxed Qi pilot:** real Topic surfaces, reversible actions, no external consequence.
3. **Human-agent study:** consented users, bounded tasks, human authority over outcomes.
4. **Consequential pilot:** limited value or credential effects, independent evaluation, rollback/dispute path.
5. **Longitudinal study:** participant turnover, artifact reuse, governance burden, and cumulative performance.

No result from an earlier stage should be described as evidence for a later-stage claim.

---

## 13. Formal model to develop

Represent a cooperative world at time `t` as:

```text
W_t = (I, P_t, S_t, A_t, R_t, G_t, V_t)
```

where:

- `I` is declared intent and completion criteria;
- `P_t` is the set of participants with identities and capabilities;
- `S_t` is observable and authoritative state, with those classes separated;
- `A_t` is the artifact set and its provenance graph;
- `R_t` is the rights and authority relation;
- `G_t` is governance, evaluation, and transition rules;
- `V_t` is resources, incentives, and consequences at stake.

For participant `p`, available affordances are:

```text
F(p, W_t) = actions observable, feasible, and permitted for p in W_t
```

An action produces a proposed transition:

```text
a(p, W_t) → ΔS*, ΔA*, claims, evidence
```

Governance and evaluation determine whether it becomes effective:

```text
E(claims, evidence, method, authority) → decision
decision + valid authority → W_(t+1)
```

This representation must be refined to model:

- partial observability;
- conflicting intents;
- uncertainty and contested evidence;
- mutable versus immutable state;
- revocation, correction, and appeals;
- privacy and selective disclosure;
- concurrent action and conflicts;
- value distribution;
- exit and the survival of prior obligations.

The formal model earns its place only if it generates testable predictions or clarifies a design decision.

---

## 14. System correspondence template

Use this template for every claimed Qi/IXO primitive.

| Field | Required entry |
|---|---|
| Theoretical primitive | The primitive from the model |
| Qi/IXO object | Exact current product/protocol name |
| Source of truth | Versioned file, schema, specification, or endpoint |
| Current lifecycle | Create, update, evaluate, complete, revoke, archive |
| Persistence | Where and for how long it persists |
| Visibility | Who or what can perceive it |
| Authority | Who can mutate or make it consequential |
| Provenance | How authorship and transition history are established |
| Evaluation | What determines acceptance or completion |
| Consequence | What effective state can change |
| Status | Implemented, partial, planned, deprecated |
| Gap | Missing requirement or unresolved semantic |
| Evidence | Test, trace, screenshot, transaction, or reproducible procedure |

The published paper should include a compressed version of this matrix and release the full matrix as a technical appendix.

---

## 15. Research records

### Claim card

```text
Claim ID:
Exact claim:
Class: [E] [X] [S] [I] [H] [N]
Scope and boundary conditions:
Supporting evidence:
Contrary/limiting evidence:
Confidence:
Chapter and argument role:
Consequence if wrong:
Status and reviewer:
```

### Source card

```text
Source ID:
Full citation and stable URL/DOI:
Primary or synthesis:
Peer-review/release status:
Study domain and sample:
Method, comparator, and outcome measures:
Findings used:
Findings not used:
Limitations:
Claim IDs supported or challenged:
```

### System evidence card

```text
System claim ID:
Object and version:
Authoritative source:
Observed behavior:
Reproduction steps:
Expected versus observed result:
Deployment scope:
Known gaps:
Last verified date:
```

### Experiment card

```text
Experiment ID:
Research question and hypothesis:
Conditions and randomization:
Task and participant population:
Primary and secondary outcomes:
Power/sample rationale:
Predefined exclusions:
Adverse-event and ethics controls:
Analysis plan:
Falsifying result:
Data/code release status:
```

---

## 16. Drafting order

Research and write in dependency order, not chapter order.

1. **Fix vocabulary and claim classes.**
2. **Complete the SwarmWorld replication card.**
3. **Build the cross-field evidence map.**
4. **Complete Qi + IXO system archaeology.**
5. **Write the formal mechanism and boundary conditions.**
6. **Run the counterargument and security review.**
7. **Specify the comparative experiments.**
8. **Draft Parts II–VI**, where the theory is established.
9. **Draft Parts VII–VIII** from verified system evidence.
10. **Draft Parts I, IX, and the conclusion** after the claims are stable.
11. **Write the abstract last.**

This prevents the opening and product sections from overpromising beyond the evidence developed later.

---

## 17. Review gates

### Gate A — Conceptual integrity

- terms are consistent and non-circular;
- stigmergy is not used as an explanation for all coordination;
- voluntariness, emergence, decentralization, and legitimacy remain distinct;
- the mechanism yields predictions or design consequences.

### Gate B — Evidence integrity

- every central empirical claim has primary support;
- recent and single-study findings are appropriately qualified;
- contrary evidence and null results are represented;
- quantitative claims include context and denominator;
- no citation is asked to support more than it actually shows.

### Gate C — System truth

- all present-tense Qi/IXO claims are verified against current sources or behavior;
- implemented, partial, and proposed capabilities are visibly separated;
- the model host is not described as the authority that establishes reality;
- trust boundaries, human gates, and revocation paths are explicit.

### Gate D — Strategic honesty

- network-effect and moat claims remain hypotheses unless measured;
- blockchain is used only where consequence requires its properties;
- alternative architectures receive a fair comparison;
- known failure modes and governance costs are treated as central.

### Gate E — Scientific usefulness

- core hypotheses are falsifiable;
- experiments include credible comparators and ablations;
- negative results would change the theory or architecture;
- data, code, definitions, and evaluation methods can be independently inspected where safe.

---

## 18. Open questions requiring system evidence

The prior narrative proposes strong correspondences, but the following must remain open until the Qi + IXO archaeology is complete:

1. Is a Topic currently a semantically bounded cooperative environment, or primarily a product/UI abstraction over room and document state?
2. Which Topic Contract fields are authoritative today: intent, outcome, scope, participants, authority, completion, and state?
3. Are Deeds implemented objects with execution and lineage semantics, or a conceptual name for work?
4. Do Flows preserve reusable execution logic together with evidence of results?
5. What exact boundary distinguishes a Flow, POD, Deed, Action, request, and Topic?
6. Which capabilities are portable and cryptographically enforceable versus application permissions?
7. What makes a claim accepted, who controls the evaluation method, and how are disputes or appeals handled?
8. Which decisions can produce credentials, payments, asset transitions, permissions, or chain state?
9. What state is ephemeral, collaborative, persistent, authoritative, and consequential in the actual stack?
10. How are stale, harmful, or superseded artifacts corrected, revoked, hidden, or retired?
11. What can a participant meaningfully refuse or exit, and what obligations survive exit?
12. What evidence already exists that Topics or reusable artifacts improve cooperation rather than organization alone?

Answers should change the paper. They should not be reverse-engineered to preserve its original claims.

---

## 19. Publication package

The research should yield six linked artifacts:

1. **White paper** — accessible narrative and central argument.
2. **Evidence dossier** — annotated primary sources and exact claims supported.
3. **Claim ledger** — machine-readable status of every substantive claim.
4. **System correspondence matrix** — present implementation versus proposed architecture.
5. **Threat and counterargument appendix** — failure modes, rival explanations, and mitigations.
6. **Experimental protocol** — preregisterable tests, metrics, datasets, and reporting rules.

The white paper is ready to draft only when the evidence dossier covers the conceptual spine and the system correspondence matrix has resolved all present-tense Qi/IXO claims.

---

## 20. Immediate research sequence

### Phase 1 — Establish the defensible core

- verify SwarmWorld at claim level;
- stabilize definitions and bridge claims;
- identify canonical and contrary evidence across the core literatures;
- create the initial claim ledger.

### Phase 2 — Establish system truth

- ingest the authoritative Qi and IXO specifications and repositories;
- trace each proposed primitive through lifecycle, authority, persistence, and consequence;
- downgrade or reframe claims that are not implemented.

### Phase 3 — Challenge the thesis

- compare orchestration, messaging, shared-state, governed, and hybrid architectures;
- complete the threat model and rival-explanation review;
- revise the primitive set and causal model.

### Phase 4 — Make it testable

- select benchmark tasks and field-pilot contexts;
- preregister hypotheses and metrics;
- instrument Qi for communication, contribution, reuse, provenance, and consequence traces;
- define publication and stop criteria.

### Phase 5 — Write and release

- draft from the verified claim ledger;
- obtain domain, systems, governance, security, and methods review;
- publish the white paper with appendices and clearly versioned evidence artifacts;
- maintain a public errata and research-results record.

---

## 21. Success condition

The research architecture has succeeded when the paper can answer, without hand-waving:

1. **What is the mechanism?**
2. **What evidence shows that it occurs?**
3. **Under what conditions does it outperform alternatives?**
4. **What makes participation voluntary and consequences legitimate?**
5. **What does Qi + IXO implement today?**
6. **What remains hypothetical?**
7. **What observation would prove the central thesis wrong or incomplete?**

The paper's strongest credible conclusion should be no larger than the evidence permits:

> Persistent shared state is a plausible and increasingly testable medium of collective intelligence. Qi + IXO can be evaluated as an architecture for making that medium legible, voluntary, governed, and consequential—but its strongest claims must be demonstrated through system evidence and comparative experiments.
