By W.H.L., GPT-5.6 Sol, Claude Sonnet 5

Gradual AGI Series #10
Formal/empirical companion to Gradual AGI as Participation: Subjects, Acts, and Consequences
Research cutoff: August 28, 2026
Publication version: v1.6.1

Abstract

This paper is the formal/empirical companion to Gradual AGI as Participation: Subjects, Acts, and Consequences. It proposes an integrated 3M Participation Taxonomy—Manifestness, Modality, and Materiality—whose three dimensions do not yet share equal measurement status, retains the realization-specific Relational Grammar of Realization Orientation (RO), Foreclosure Position (FP), and Actor Dependence (AD), and specifies a Responsive Floor Test. The three analytical objects also do not have equal empirical status in this study: only Modality and RO/FP/AD were prospectively frozen in outcome-blind Packet A; Manifestness and Materiality are secondary descriptive applications to preserved acts; and the aggregate Responsive Floor Test is not adjudicated.

The primary domain is U.S. AI-related data-center development. A bounded sponsor/project-announcement frame was frozen at , followed by a ten-case instrument-development pilot and a seeded 21-row confirmatory draw. At the August 28, 2026 cutoff, 14 cases were mature and seven right-censored; the mature set contained 11 Present, two SearchedNotFound, and one Unknown. The denominator crosswalk is partly reconstructed: 11 mappings were directly recovered and 38 mechanically reconstructed under a rule that reproduces all recovered anchors. A complete 1,176-pair identity audit identified one physical-campus collision, F1-043/F1-044; the two rows fall in pilot and confirmatory samples respectively, so the campus is not duplicated in confirmatory estimates.

The strongest executed result is structural rather than causal: Participation Mode does not encode consequence. The same frozen M5 contestatory/claim-making Mode occurs in a bounded Null, in informational/procedural consequence without attributed realization conversion, and in configurations that reach realization-level conditioning. In the executed single-coder, outcome-blind coding, RO/FP/AD preserve distinctions not captured by visible Mode, form, or project stage. These observations are hypothesis-generating: because the complete matching pre-adjudication reliability pair is not reconstructable, independent reproducibility remains untested and no instrument-validation claim is made.

Port Washington supplies an L3 documented association sequence in the institutional record: participants lacking binding authority enter a municipal process, and the City later attributes zoning and sustainability conditioning to community input. This establishes documentary sequence and institutional attribution, not counterfactual contribution, necessity, sufficiency, or a general mechanism. Lancaster supplies weaker L2 configuration evidence. The aggregate/domain Responsive Floor verdict remains NotAdjudicated because the silent comparator is too small and the realization environments too heterogeneous.

The secondary 3M pass yields 16/16 Explicit primary acts and Materiality of 10 Intangible, 5 Mixed, and 1 Tangible. These counts are an observation-strategy diagnostic, not a prevalence estimate or validation of the full taxonomy: public records systematically favor overt, document-producing participation. The temporal extension likewise does not identify a positive calendar-2026 resistance-frequency effect; uncertainty is too wide to establish either an increase or its absence. Formally, Frame One supports only a typed partial-transition specification for future research, not a universal Participation scalar, type effect, or additive coefficient.

The paper therefore contributes: (1) a proposed realization-centered 3M Participation Taxonomy whose dimensions do not yet share equal measurement status; (2) outcome-blind observations showing that the Relational Grammar preserves distinctions not captured by visible Participation Mode in this coding, with reproducibility untested; (3) a documented non-veto association sequence with counterfactual contribution unidentified; and (4) a confirmatory-design methodology that separates sampling and coding discipline from inferential operability.

Keywords: Gradual AGI; participation; 3M Taxonomy; Manifestness; Modality; Materiality; Participation Mode; Relational Grammar; AI governance; data centers; realization; Responsive Floor; non-veto participation.

Evidence and source note

This technical companion uses frozen Packet A/Packet B records for case-level claims; the full 21-case evidence/source register is provided separately with this publication package as Supplement S1. Appendices preserve the codebook, denominator, Packet B protocol, reliability provenance, temporal specifications, pilot status, and sensitivities. No denominator, draw, Packet-A/Packet-B code, consequence code, linkage grade, or realization adjudication is changed in v1.6 or v1.6.1. The integrated 3M Taxonomy is a proposed framework: Modality is inherited from the frozen Act Form field, while Manifestness and Materiality remain secondary descriptive applications requiring prospective replication.

1. Introduction

Gradual AGI refers here to the process by which increasingly capable AI becomes socially realized through ordinary institutional, material, economic, and governance arrangements rather than only through a discrete capability threshold. This paper examines where heterogeneous human participation enters that realization process. Although AI infrastructure supplies the empirical domain, the analytical architecture is intended to travel to other realization processes in which human acts interact with institutional and material outcomes.

Gradual AGI as Participation separated three objects that ordinary accounts of participation often collapse: the Participatory Subject, the Participatory Act, and the Participatory Consequence. It also proposed a three-part Relational Grammar—Realization Orientation, Foreclosure Position, and Actor Dependence—and asked how participation might enter realized conditions even when participants do not themselves hold formal veto or binding authority. This companion turns that question into a named Responsive Floor Test while keeping documentary linkage separate from causal contribution.

The conceptual paper deliberately did not complete the taxonomy of participatory forms. It fixed what a Participatory Act is—observable intentional action related to an identifiable realization—and handed the empirical companion the harder task of determining how heterogeneous participation should be represented in practice. That task is central here.

The starting claim is simple but consequential:

Participation can be explicit or implicit, can appear in different forms, and can be tangible or intangible. No participation type carries a consequence by definition. The relation between participatory type and participatory consequence is contingent, and identifying where and how those transitions occur is a principal empirical problem.

This paper therefore treats Participation neither as a binary exposure nor as a scalar quantity of “voice.” It proposes a multidimensional Participation Taxonomy inside the Act layer. A participatory act can be classified by the 3M Participation Taxonomy: Manifestness (explicit/implicit), Modality (one of eight Participation Modes, M1–M8), and Materiality (tangible/intangible/mixed). Those taxonomic dimensions are then distinguished from the Relational Grammar—RO, FP, and AD—which describes how the act is situated relative to the realization it addresses. Institutional recipient, authority, channel, procedure, timing, and related fields are kept separate again because they condition conversion rather than define the participation type itself.

The empirical object is thus not merely whether participation occurred. It is the contingent sequence:

with neither arrow assumed to exist, to be monotone, or to be the same across types or institutional settings.

The empirical design has a deliberately mixed epistemic structure. Frame construction and sample selection are confirmatory design objects: the project denominator is outcome-independent, the pilot and confirmatory draws use frozen seeds, Packet A is coded before Packet B is opened, and the denominator is not redrawn because a case proves inconvenient. But the mature cells are too sparse for a confirmatory aggregate participation-effect estimate. In addition, only Participation Mode (recorded under the frozen Packet-A field name Act Form) and RO/FP/AD were frozen as act-level confirmatory fields. Manifestness and Materiality are proposed as the other two dimensions and applied only secondarily to preserved acts; they are not retroactively treated as frozen Packet-A variables or as validated measures.

The primary domain is U.S. AI-related data-center development. It is not claimed to represent participation generally. It is analytically useful because AI infrastructure must become ordinary material and institutional arrangements: land, electricity, water, generation, permits, utility tariffs, tax structures, construction, noise controls, reporting duties, and local bargaining. Residents, utilities, developers, governments, regulators, courts, and organized groups enter different parts of that realization path with different forms of participation, authority, and standing. The same domain also supplies hard negatives: projects with no recovered qualifying focal act, explicit acts with no detected consequence, formal procedures without realization conversion, and authoritative acts that enable rather than resist a project.

1.1 Research questions

The paper asks five questions.

1. What taxonomy is needed to represent heterogeneous Participation without collapsing it into volume, visibility, or a single behavioral category?
2. Do Participation Mode and the Relational Grammar preserve distinctions that ordinary descriptors such as channel, count, formality, or project stage would lose?
3. How do different participation types and participatory signatures relate to Participatory Consequences, and where do those transitions stop?
4. Responsive Floor Test: what documentary association sequences link pre-specified Participatory Configurations to later realization conditions, including through participants lacking binding authority, and when is the evidence sufficient for a domain-level verdict?
5. What formal representation, if any, is warranted after empirical contact?

1.2 Evidence-status box

The design separates confirmatory discipline from inferential operability. The outcome-independent denominator, seeded draws, Packet-A-before-Packet-B ordering, censoring rules, and anti-rescue rules were fixed before outcome analysis. The final confirmatory instrument, however, was frozen after an instrument-development pilot, and the surviving pilot archive does not preserve a complete repair ledger or final-codebook pilot recoding. Confirmatory discipline therefore describes the executed post-pilot design; it does not imply that the instrument was fully pre-specified before pilot development, adequately powered, or validated.

Object	What the present evidence supports	What it does not support
3M Participation Taxonomy	proposed integrated act-level taxonomy; Modality prospectively exercised; Manifestness/Materiality applied descriptively	full 3M validation, prevalence, or type-effect estimates
Relational Grammar (RO/FP/AD)	distinctions preserved in the executed single-coder, outcome-blind coding; hypothesis-generating	inter-coder reproducibility or instrument validation
M5 contingency	same frozen Mode ends in different consequence profiles	explanation of which condition causes conversion or relative effectiveness
Non-veto sequence	Port Washington L3 institutional-record association sequence	counterfactual contribution, necessity, sufficiency, or general mechanism
Responsive Floor Test	association component exercised; aggregate comparator inspected	positive or negative domain-level verdict
Temporal extension	no positive calendar-2026 effect identified in the fitted sparse models	proof of no increase, decline, or maturation law

Two methodological diagnostics are equally important. First, the mature silent comparator contains only two SearchedNotFound cases, far below the frozen precision gate. Second, the public-record strategy recovers overt document-producing participation much better than implicit, behavioral, productive, or adaptive participation. The 16/16 Explicit secondary occupancy is therefore evidence about observation opportunity, not population prevalence.

1.3 Contribution

The paper makes four bounded contributions.

1. It proposes the 3M Participation Taxonomy as a realization-centered act-level taxonomy, while explicitly separating prospective exercise of Modality from secondary descriptive application of Manifestness and Materiality.
2. It shows, in the executed single-coder, outcome-blind coding, that RO/FP/AD preserve distinctions not captured by visible Participation Mode, project stage, or ordinary form descriptors. These observations are hypothesis-generating; independent reproducibility remains untested.
3. It establishes the structural negative finding that M5 does not encode consequence, and documents one L3 non-veto association sequence in Port Washington without converting institutional attribution into causal identification.
4. It supplies a reusable empirical architecture for participation research: outcome-independent framing, act-before-outcome coding, explicit missingness, precision gates, source-coverage diagnostics, and no-rescue rules.

The paper does not validate a general participation-effect model, a reproducible measurement instrument, the full 3M Taxonomy, or a domain-level Responsive Floor.

1.4 Historical position and contribution boundary

The paper does not claim to invent multidimensional participation classification. Arnstein (1969) and Pateman (1970) link participation to power and authority; Rowe and Frewer (2005) classify engagement mechanisms; Fung’s (2006) democracy cube is the closest classic multidimensional institutional design; Ekman and Amnå (2012), Villi and Matikainen (2016), and Theocharis and van Deth (2016/2018) provide direct antecedents for manifest/implicit participation and participation modes; material-participation research foregrounds embodied and material arrangements (Marres 2015; Rydin and Natarajan 2016); and participatory-AI research has already imported participation questions into AI design and governance (Birhane et al. 2022; Delgado et al. 2023; Nesta 2021; OECD 2026).

Relative to the closest predecessors, the distinction is compact. Unlike Arnstein’s ladder, 3M is non-ordinal. Unlike Fung’s democracy cube, it classifies the participatory act rather than the institutional design of the participatory mechanism. Unlike manifest/latent or explicit/implicit distinctions, it treats Manifestness alongside Modality and Materiality as simultaneous act-level dimensions.

The contribution is narrower and more specific: an act-level, realization-centered integration of Manifestness, Modality, and Materiality; its analytical separation from a realization-specific Relational Grammar; an explicit contingency rule preventing type from determining consequence by definition; and a Responsive Floor Test that separates documentary association from domain-level adjudication. Institutional recipient, authority, procedure, timing, and channel remain conditioning variables rather than taxonomy dimensions.

The term Relational Grammar itself is not claimed as original; it has a separate history in linguistics. The present contribution is the RO–FP–AD construction for Participatory Acts relative to an identifiable realization. Likewise, context-dependent participation effects are not a new discovery. The novelty lies in the separation and integration of taxonomy, relational position, consequence, and realization under a common anti-deterministic architecture.

A compact summary is:

The novelty is not participation classified; it is participation classified relative to realization and kept analytically separate from consequence.

2. Framework, 3M Participation Taxonomy, and Adjudication Criteria

This paper inherits the Subject → Act → Consequence architecture from #9 and adds the 3M Participation Taxonomy that the conceptual paper deliberately left incomplete. Appendix A consolidates notation and formal objects; Appendix B provides the taxonomy and act-level codebook.

2.1 Inherited realization mapping

The inherited Gradual AGI realization mapping remains:

where  is capability potential,  group-specific exposure, and  the relevant governing arrangement. Exposure remains structured as:

with reach, susceptibility, and recourse kept separate. CFS separately retains actor aggregation through . Nothing in #10 inserts Participation into this mapping by assumption.

The Participation question begins at a narrower point: whether heterogeneous human participatory configurations can become relevant to the state on which realization operates, including through subjects who do not themselves hold binding authority.

2.2 Participatory Subject and Participatory Act

A Participatory Subject is a human person or human-organized collective capable of intentional action with respect to an identifiable realization. Affectedness alone is insufficient.

A Participatory Act is an observable intentional act related to an identifiable realization object. The act must therefore be attributable, realization-related, and observable in evidence. The taxonomy below classifies qualifying acts; it does not broaden the inclusion gate to passive exposure, unexpressed attitudes, or behavior whose intentional relation to the realization cannot be established.

2.3 The 3M Participation Taxonomy

Definition — 3M Participation Taxonomy. The 3M Participation Taxonomy is a multidimensional classification of a qualifying Participatory Act along three descriptive dimensions: Manifestness, Modality, and Materiality. Modality takes one of eight non-ordinal Participation Modes, M1–M8. The taxonomy is nested inside the Act layer and does not replace the Subject → Act → Consequence architecture.

For act , define the taxonomic type:

where  is Manifestness,  is Modality/Participation Mode, and  is Materiality.

The dimensions answer different questions:

• Manifestness: how overtly is the participatory relation expressed?
• Modality: what does the subject do?
• Materiality: how is the act instantiated?

The short label 3M Taxonomy refers to this Manifestness–Modality–Materiality construction. The originality claim concerns their realization-centered integration at the act level, not the historical origin of each individual concept.

2.3.1 Manifestness: explicit and implicit Participation

Explicit Participation is a qualifying act that overtly presents itself as addressing an identifiable realization—supporting it, opposing it, seeking to condition or redirect it, reporting on it, requesting institutional action, authorizing it, or otherwise making the participatory relation manifest.

Examples include a petition, public comment, formal objection, lawsuit, vote, hearing intervention, regulatory filing, project-specific authorization, or an express declaration of support, refusal, or demand.

Implicit Participation is a qualifying intentional act whose relation to an identifiable realization is established through behavior, practice, contribution, adaptation, or organization without the subject overtly presenting the act as a participatory intervention.

Examples can include intentional adoption or non-adoption, exit, contribution, production, workflow reorganization, training, refusal, or other behavior undertaken in relation to the realization without an accompanying explicit claim of “participation.”

Implicit does not mean unconscious, merely inferred from affectedness, or unobservable. Intentional relation must still be supportable from the record. Passive exposure, latent preference, or an unexpressed private attitude is not converted into Participation merely because it might matter causally.

A corresponding sentiment boundary is fixed here. A private, unexpressed sentiment is not a Participatory Act. An intentionally public or socially accessible expression of sentiment concerning AI realization is participation, whether the sentiment is distrust, worry, fear, enthusiasm, approval, criticism, humor, parody, or mockery. Such content normally enters M1 Expressive / Discursive unless the operative act is better classified as structured feedback, deliberation, contestation, or another Mode. Participation therefore does not require institutional receipt or observable consequence.

2.3.2 Modality: eight Participation Modes

Modality records the operative kind of participatory action. Its categorical values are the eight Participation Modes:

Code	Participation Mode	Core content
M1	Expressive / Discursive	public expression, speech, publication, commentary, joke, meme, parody, mockery, praise, criticism, or discourse
M2	Evaluative / Feedback	evaluation, review, feedback, assessment, structured reporting, poll/survey response
M3	Behavioral / Choice	adoption/non-adoption, exit, use/non-use, purchase or other realization-relevant choice
M4	Deliberative / Consultative	consultation, deliberation, hearing, advisory participation, structured discussion
M5	Contestatory / Claim-making	petition, objection, protest, challenge, appeal, lawsuit, demand
M6	Productive / Co-creative	contribution, collaboration, dataset/content production, co-creation
M7	Adaptive / Reorganizational	training, institutional adaptation, role reorganization, workflow or capacity change
M8	Authoritative / Constitutive	approval, authorization, rule-making, or other act undertaken through binding/constitutive authority

The categories classify the act, not the actor, topic, sentiment, channel, or consequence. The same subject can move among Modes over time, and one episode can contain multiple subjects, acts, and Modes.

2.3.3 Materiality: tangible, intangible, and mixed Participation

Intangible Participation operates primarily through expression, information, evaluation, interpretation, symbolic positioning, or other communicative/knowledge-bearing action. Its immediate participatory substance is not a direct change in a material, behavioral, resource, or institutional state.

Tangible Participation is instantiated through action that directly changes behavior, an artifact, resource allocation, organizational practice, legal/institutional state, or a material component of realization.

Mixed Participation combines both in a way that cannot be cleanly separated at the act level—for example, an act whose communicative content and constitutive institutional operation are jointly part of what the subject does.

Materiality concerns the instantiation of the act, not the magnitude of its later consequence. An intangible petition can have major realization consequences; a tangible act can fail to produce the consequence sought. Tangible therefore does not mean effective, and intangible does not mean weak.

2.3.4 Measurement status in the present study

The 3M Participation Taxonomy is a proposed integrated taxonomy, not a three-dimensionally validated instrument in Frame One.

Dimension	Present-study status
Manifestness	secondary descriptive application to preserved acts; no prospective reliability/effect claim
Modality / Participation Mode	prospectively frozen in Packet A under Act Form; primary-signature occupancy reported
Materiality	secondary descriptive application to preserved acts; no prospective reliability/effect claim
RO / FP / AD	prospectively frozen Packet-A fields; distinctions observed in executed coding; independent reproducibility untested

The secondary pass assigns all 16 frozen primary acts as Explicit and Materiality as 10 Intangible, 5 Mixed, and 1 Tangible. Appendix J gives the full matrix. These counts may diagnose source coverage and illustrate prospective coding rules; they may not be used as prevalence estimates, effect comparisons, or evidence that the full 3M instrument is reliable.

Full3MTaxonomyValidation_FrameOne = NotExercised

2.3.5 Non-sample exposition: MIT’s 2026 AI-use review as a multi-Mode configuration

This example is not Frame One evidence. MIT’s 2026 Ad Hoc Committee episode is used only to clarify the unit rule: structured assessments can be M2 Evaluative/Feedback, reciprocal consultation M4 Deliberative/Consultative, later binding policy adoption M8 Authoritative/Constitutive, and later workflow or curriculum reorganization M7 Adaptive/Reorganizational. Those are different subject-act pairs inside one episode; later implementation must not be back-propagated into the earlier report or consultation.

The taxonomy therefore classifies the specific act by the specific Participatory Subject, not an entire episode. The MIT example is expository only and is excluded from every Frame One count, contrast, and adjudication.

2.4 Relational Grammar and the participatory signature

The taxonomy answers what kind of participation is occurring. The Relational Grammar answers how that act is situated relative to the realization it addresses:

• Realization Orientation (RO): supportive, conditional/redirective, oppositional, or reporting/nondirective relative to the specific referent;
• Foreclosure Position (FP): open, constrained, foreclosed/remedial, or unresolved relative to the meaningful revisability of that referent;
• Actor Dependence (AD): how much evidentiary, experiential, or standing-related content would be lost if another actor were substituted.

RO is not general AI sentiment. FP is not generic project stage. AD is not formal power.

Taxonomy and Grammar together form a participatory signature:

The distinction is essential. Two acts can share the same Participation Mode and still differ materially in orientation, revisability, or actor dependence. Conversely, similar Grammar states can be instantiated through different forms or degrees of manifestness/materiality.

Proposition A — Conditional Non-Fungibility states that visible equivalence in form, channel, or volume is insufficient to establish act equivalence where RO, FP, or AD differs materially. Frame One can only exercise this proposition inside the executed coding; whether independent coders reproduce the same distinctions remains a separate, unresolved burden.

2.5 Participatory Consequence and the contingency principle

A Participatory Consequence is a change, persistence, or bounded absence of detected change associated with an observed act or configuration. The seven non-exclusive classes are Informational, Procedural, Institutional/return-path, Subject-readiness, Distributional, Realization-level, and Null/no detected change.

The taxonomy does not assign any consequence in advance. Formally:

and more generally:

The empirical object is a contingent relation rather than a deterministic mapping. A type can be associated with different consequence profiles under different recipients, authority structures, procedures, timing, bargaining conditions, and other realization constraints.

The architecture therefore permits:

That expression is not a probabilistic model. It states that each transition must be observed rather than inferred from the participation type itself.

2.6 Proposition B as an operability result

Participatory Role Plasticity requires an observable human-task automation transition and longitudinal observation of the associated human role. Frame One does not instantiate that antecedent.

Because Proposition B was inherited prospectively from #9, deleting it after observing the domain would itself be outcome-conditioned selection. It therefore remains only in the proposition-status ledger and future-research requirements:

PropositionB = UntestedInFrameOne.

No null, support, or effect claim is attached to it.

2.7 Responsive Floor Test and Proposition C

A Responsive Floor is the theoretical state in which human Participatory Configurations contribute to differences in realized Floor conditions, including through subjects lacking formal veto or binding authority.

Definition — Responsive Floor Test. The test separates two burdens:

1. Association-sequence burden: does the institutional record contain a traceable sequence linking a pre-specified Participatory Configuration, an institutional process, and a later binding or material realization condition?
2. Aggregate/domain burden: does an outcome-independent comparison with adequate precision and comparability support classifying the domain as responsive to differences in Participatory Configuration?

In the data-center domain, the first burden would be met when resident participation enters a municipal decision process and the institution later attributes a project condition to that input. The second would require sufficiently comparable projects showing that such configurations systematically differentiate realization rather than merely co-occur with it.

The two burdens must not be collapsed. Documentary association can be present while counterfactual contribution and the domain-level verdict remain unidentified.

For Frame One:

ResponsiveFloorTest_AssociationSequence = DocumentedAssociationSequenceObserved

ResponsiveFloorTest_Aggregate = NotAdjudicated

ResponsiveFloor_FrameOne = NotAdjudicated.

Appendix A.6 states the adjudication criteria; Appendix A.7 states the additional burden required for mechanism identification.

3. Design, Denominator, Coding, Reliability, and Operability

The design separates project membership, act coding, consequence coding, and realization coding so that later outcomes cannot determine who enters the frame or how an earlier act is described. Detailed protocols are preserved in Appendices C–I.

3.1 Frame, denominator, and sampling

The executed denominator is a sponsor/project-announcement frame frozen at . The randomized unit is the announcement row, not necessarily a permanently unique physical campus. The original full crosswalk does not survive: 11 mappings were directly recovered and 38 mechanically reconstructed using a rule that reproduces every recovered anchor. The original written ordering instruction was not recovered verbatim. The denominator is therefore canonical for this project but partly reconstructed, and every confirmatory claim inherits that provenance limitation.

A complete  pair identity audit identified one physical-campus collision, F1-043/F1-044. F1-043 is pilot-only and F1-044 confirmatory, so the campus is not duplicated in confirmatory estimates. F1-044 remains separately covered by leave-one-out sensitivity.

The AI-relevance audit finds 20/21 explicit project-side passes and one Moderate-confidence contextual/public case, F1-013; no confirmatory case relies on governance-side AI relevance alone. Excluding F1-013 leaves all headline conclusions unchanged.

The pilot uses seed 20260828, , for instrument development. The confirmatory draw uses seed 20260829, , and no redraw or substitution is permitted. The final confirmatory instrument was post-pilot frozen before confirmatory coding. The surviving archive does not preserve a complete pilot repair ledger or final-codebook pilot result set, so the paper does not claim that the exact confirmatory instrument was pre-specified before pilot development or that pilot/confirmatory consistency can be retrospectively demonstrated. The pilot served instrument development only and is not claimed as validation.

At the August 28, 2026 cutoff:

3.2 Packet A and Packet B separation

Packet A is fixed before Packet B is opened. It records ParticipationEvidence, qualifying acts, frozen Act Form / Participation Mode, RO, FP, AD, recipient/channel, procedure entry, authority/revisability, AI-semantic fields, scope, and provenance. Manifestness and Materiality were not frozen Packet-A variables and enter only through the secondary descriptive pass.

ParticipationEvidence has three states: Present, SearchedNotFound, and Unknown. A mature SearchedNotFound requires the standardized public-record source audit specified in Appendix C; it means no qualifying focal act was recovered under that observation strategy, not that no participation occurred.

Packet B separately codes project realization and seven non-exclusive Participatory Consequence classes: Informational, Procedural, Institutional/Return Path, Subject Readiness, Distributional, Realization Level, and Null/No Detected Change. A bounded Null requires an observed act plus a complete negative consequence search. Linkage grades L0–L4 describe evidentiary support, not causal effect size; L3 is direct documentary attribution and L4 adds independent corroboration.

3.3 Pre-specified contrasts and precision

Contrasts A–D examine procedural access, channel control, route to binding authority, and referent-specific revisability using frozen Packet-A predictors. Contrast B is not exercised because its comparator is absent. Contrast E compares project realization for Present versus SearchedNotFound, but its frozen inferential gate requires at least 20 determinate projects in each binary cell plus a Wilson half-width threshold. With only two mature SearchedNotFound cases, Contrast E is descriptive only; its table is moved to Appendix K so the main text does not invite inferential reading.

3.4 Temporal/context extension

Calendar 2026, project age, and semantic/context change are kept separate. Focal resistance events are exposure-adjusted by project-months; formal-entry and downstream-adaptation models are exploratory. Survey sentiment, news framing, social discourse, and observed participation remain separate evidence classes. The news/social protocols were not executed as quantitative result corpora.

3.5 Central measurement limitation: reliability

The registered reliability statistic cannot be reconstructed. One complete blind holdout return survives, but the complete matching pre-adjudication pair does not survive with sufficient coder/version/target/timing provenance.

Accordingly, the current RO/FP/AD result is limited to this statement:

In the executed single-coder, outcome-blind coding, RO, FP, and AD preserve distinctions; the surviving archive cannot establish whether independent coders reproduce those distinctions. These observations are hypothesis-generating, not confirmed measurement.

No inter-coder reliability, reliability failure, zero-reliability, or instrument-validation claim is made. The Grammar should therefore be read as a proposed instrument with hypothesis-generating discrimination evidence in this coding. Manifestness and Materiality require a separate future taxonomy-complete reliability study.

PropositionA_Discrimination = ObservedInExecutedCoding_HypothesisGenerating

PropositionA_Reliability = NotReconstructable_ArchivalLoss

InterCoderReproducibilityClaim = None

3.6 Operability summary

Object	Frame One status
3M Taxonomy	proposed; Modality prospectively exercised; Manifestness/Materiality secondary descriptive
Proposition A	discrimination observed in executed single-coder coding; hypothesis-generating; reproducibility untested
Proposition B	automation antecedent absent; UntestedInFrameOne
Responsive Floor Test	association sequence observed; aggregate/domain NotAdjudicated
Contrast B	not exercised; comparator absent
Contrast E	descriptive only; precision gate unmet
2026 frequency	positive calendar effect not identified; wide uncertainty
capability-escalation fear	not identified as focal explanation

The anti-rescue rule is unchanged: outcome inspection does not alter frame membership, windows, missingness, coding definitions, contrasts, precision gates, or censoring.

4. Participation Taxonomy and Relational Grammar

Case-level source provenance for this section is reproduced in Supplement S1; the taxonomy and complete act-level coding definitions are in Appendix B.

The first empirical task is now stated at the correct level: what kinds of Participation does the instrument distinguish, which of those kinds are actually exercised in Frame One, and what information is lost if acts are collapsed to one visible label? The Participation Taxonomy supplies Manifestness, Participation Mode, and Materiality; the Relational Grammar then differentiates acts by orientation, revisability, and Actor Dependence.

The executed design directly tests only the frozen Act-Form and Grammar fields. Manifestness and Materiality organize the interpretation of cases but are not promoted into retroactive confirmatory codes. This asymmetry is itself informative: a public-record data-center frame recovers explicit petitions, comments, filings, hearings, protests, and authorizations much more readily than implicit behavioral, productive, or adaptive participation. The taxonomy therefore reveals not only variation in the cases but also variation in what the domain can observe.

Sections 4.1–4.3 establish taxonomic coverage and its limits. Sections 4.4–4.8 test whether visible type and ordinary descriptors are sufficient. Section 4.9 returns to the reliability gate.

4.1 The empirical population for the Grammar test

The Grammar test uses the 16 confirmatory projects for which Packet A had documented at least one qualifying Participatory Act by the August 28, 2026 cutoff. This is a different analysis population from the 11 mature consequence-assessable cases used in Section 5.

The distinction follows directly from the Packet-A/Packet-B separation. Once an act has occurred and its contemporaneous evidence is sufficient, its Participation Mode, Realization Orientation, Foreclosure Position, and Actor Dependence can be coded even when the project’s twelve-month consequence window has not yet matured. Right-censoring therefore blocks final no-act and outcome claims; it does not erase an already observed act.

The 16 Present projects are:

F1-008, F1-009, F1-013, F1-016, F1-017, F1-021, F1-025, F1-027, F1-032, F1-033, F1-038, F1-039, F1-042, F1-044, F1-047, and F1-048.

Two further mature projects are SearchedNotFound, and three projects are Unknown. They have no act-level Grammar assignment for the focal window. They are therefore excluded from Grammar occupancy rather than assigned synthetic zero or default values.

For a compact occupancy description, one frozen primary signature is taken from each of the 16 Present projects. This is only a descriptive indexing device. It is not an act-frequency estimate, because several projects contain multiple qualifying acts and because later within-project acts are retained rather than collapsed into the primary signature.

Table 1. Primary-signature occupancy among the 16 Present confirmatory projects

Field	Values observed	Primary-signature count
Participation Mode	M5 contestatory / claim-making	13
	M4 deliberative / consultative	1
	M1 expressive / discursive	1
	M8 authoritative / constitutive	1
Realization Orientation	RO-C conditional / redirective	10
	RO-O oppositional	4
	RO-R reporting / nondirective	1
	RO-S supportive	1
Foreclosure Position	FP-O open	10
	FP-F foreclosed / remedial	4
	FP-C constrained	2
Actor Dependence	AD-L low	10
	AD-M moderate	3
	AD-H high	3

These counts index one frozen primary signature per Present project. They are not act-frequency estimates, because projects may contain multiple qualifying acts; the table describes only the analytic space exercised by the frozen primary-signature instrument.

The concentration in Participation Mode is analytically useful. Thirteen of the 16 primary signatures are internally coded M5, contestatory or claim-making. If visible form already carried the relevant structure, these acts should be difficult to distinguish further. Instead the dominant form contains substantive differences in target, timing, evidentiary dependence, and requested change. The Grammar’s strongest empirical test therefore occurs within, not between, visible Modes.

The occupancy table also establishes a weaker but necessary point: none of the three Grammar relations is constant in Frame One. In the frozen primary analysis, all four RO values appear; all three resolved FP states appear; and low, moderate, and high AD all appear. That is not yet evidence of nonredundancy by itself. A field can vary and still duplicate another observable. The stronger evidence comes from controlled contrasts.

One part of the breadth statement is sensitive to the reconstructed F1-044 row. Excluding F1-044 in the leave-one-out sensitivity removes the sole primary-signature M8 / RO-S observation, leaving RO-C, RO-O, and RO-R. It does not remove the same-form/different-RO contrasts, within-project FP variation, or AD contrasts below. The nonredundancy result therefore does not depend on the existence of the sole supportive primary signature.

Table 2. Secondary descriptive 3M occupancy among the 16 frozen primary acts

Dimension	Descriptive occupancy	Measurement status
Manifestness — Explicit	16	secondary descriptive
Manifestness — Implicit	0	under-observed; not a comparator
Materiality — Intangible	10	secondary descriptive
Materiality — Mixed	5	secondary descriptive
Materiality — Tangible	1	secondary descriptive

These values were assigned after the original Packet-A freeze from the preserved act record. They are reported to expose coverage and organize examples, not as confirmatory frequencies, prevalence estimates, or effect strata. Appendix J reproduces the act-level matrix.

4.2 Manifestness: the observed primary-act set is entirely explicit

The secondary descriptive 3M pass classifies all 16 frozen primary acts as Explicit and none as Implicit. That result is best interpreted as an observation-strategy diagnostic, not a substantive estimate of participation prevalence.

The public-record design preferentially recovers petitions, public comments, testimony, protests, regulatory motions, lawsuits, and project-specific authorizations—acts that overtly externalize a claim, evaluation, report, or authoritative action. It is much weaker for intentional non-adoption, exit, quiet refusal, contribution, adaptation, or reorganization that may be realization-relevant without presenting itself publicly as participation.

Accordingly:

ImplicitParticipation_FrameOne = UnderObserved_NotTestedAsComparator

and no explicit-versus-implicit consequence rate is reported. A future study that lacks a credible observation channel for implicit participation should code that mode as UnderObserved or Unresolved, not as substantively absent.

4.3 Materiality: secondary description without effect inference

The secondary descriptive pass classifies the 16 frozen primary acts as 10 Intangible, 5 Mixed, and 1 Tangible. Materiality classifies the act’s immediate instantiation, not the magnitude of its consequence.

The coding rule is intentionally conservative. ProcedureEntry=Yes does not automatically make an act Mixed. A hearing comment or testimony can remain Intangible when the procedure is the channel through which communicative content travels. A formal motion, legislative amendment, lawsuit, or embodied protest is Mixed when the focal act itself inseparably combines claim/information with a legally operative, procedural, or embodied component. F1-044’s project-specific water authorization is Tangible because the authoritative act directly changes the institutional/material conditions of water-line implementation.

The useful result is therefore representational rather than comparative: the framework can distinguish communicative, mixed, and constitutive acts without assuming that Tangible means stronger or that Intangible means weak. No Materiality-specific effect estimate is attempted, and these placements require prospective reliability testing before use as confirmatory variables.

4.4 Participation Mode is not the Grammar

The empirical records support the distinction made in #9 between what an act visibly is and how it is situated in a realization process.

Reader-facing Participation Mode describes the observable action. The research archive retains M1–M8 identifiers, but those identifiers are not a fourth Grammar dimension. In the confirmatory frame the primary signatures occupy four visible Modes:

• expressive / discursive;
• deliberative / consultative;
• contestatory / claim-making;
• authoritative / constitutive.

The three Grammar relations ask different questions:

• RO: what relation does the act take toward the realization it addresses?
• FP: how open is the relevant realization choice when the act occurs?
• AD: what substantive content would be lost if another actor were substituted?

The distinction matters because the same broad form repeatedly carries different answers.

A petition can seek pause, redesign, or permanent opposition. A formal comment can report information, request conditions, or oppose a decision. A complaint can be highly actor-dependent because it supplies first-person exposure evidence, or weakly actor-dependent because another similarly situated intervenor could reproduce the same legal or technical claim. An authoritative institutional act can enable realization rather than restrict it.

The data therefore reject a simple shortcut:

The rest of this section shows the failure separately for each relation.

4.5 Realization Orientation (RO): similar acts can point in different directions

Realization Orientation produces the clearest same-form contrast.

4.5.1 Two petitions, two orientations in El Paso

F1-038 contains two public petition acts concerning the Meta El Paso project.

Jessica Burnett’s May 3, 2026 petition sought a pause in construction and a reopening or redirection of the decision trajectory. It was coded:

Carlos Perez’s May 22 petition sought to stop the development. Its visible form remained broadly the same—public petition / contestatory claim-making—but its orientation was oppositional:

Calling both simply “petitions against the data center” would erase the distinction between a demand to pause and alter realization and a demand to prevent it.

4.5.2 Port Washington repeats the distinction

F1-048 supplies the same result in a separate project. Shane Schmutzler’s August 19, 2025 annexation-hearing intervention sought redirection through referendum or additional public control and was coded RO-C. Joshua Taylor’s later “No AI Data Center” petition and the November formal TID opposition were coded RO-O.

Again, the broad form—contestatory public participation—does not determine the orientation.

4.5.3 Formal participation is not inherently restrictive

The contrast is not limited to varieties of opposition. F1-044 is an authoritative institutional act by Hamby Water Supply Corporation concerning project-specific water infrastructure for Vantage Frontier. It is internally coded M8, authoritative / constitutive, but its orientation is:

The act enables a project-specific water line and therefore supports realization. Formal institutional action cannot be assumed to mean restriction, contestation, or governance “against” a project.

This supportive observation is sensitive to F1-044 retention. Excluding F1-044 removes the sole primary-signature M8 / RO-S case and the direct-authority comparator, but it does not remove the same-Mode/different-RO contrasts elsewhere or the Port Washington non-veto association sequence. The supportive category is therefore illustrative but not load-bearing for Proposition A.

Together these cases show why the series’ participation concept cannot be reduced to resistance. The confirmatory frame contains RO-S, RO-C, RO-O, and RO-R. Orientation has to be read from the act’s relation to the realization object, not inferred from whether the act looks formal, political, contentious, or public.

4.6 Foreclosure Position (FP): project stage is too coarse

Foreclosure Position is the dimension most damaged by project-level coding.

A project can be physically under construction while a later expansion, operating condition, infrastructure decision, tax arrangement, permit, or second-order governance rule remains open. Conversely, an early-looking public act can occur after the particular decision it challenges has already been substantially settled. The empirical unit therefore has to be the act–referent pair, not the project as a whole.

4.6.1 Lancaster: FP-F → FP-C → FP-O inside one project

F1-025 provides the strongest within-project demonstration.

The primary July 22, 2025 stop-project intervention occurred after major first-phase commitments and was coded:

A later formal zoning appeal targeted a still-reviewable but narrowed legal route and was coded:

Still later, July 1, 2026 public comment concerned a later project phase entering an open Planning Commission review and was coded:

Thus one project contains:

without any contradiction. The arrow is not a claim that the original project somehow became less developed over time. The acts address different realization referents. The first-phase stop demand is substantially remedial; the zoning appeal addresses a constrained legal path; the later-phase review still contains materially different available choices.

A single variable such as “project under construction” or “late-stage project” would collapse all three.

4.6.2 Hyperion: different institutional trajectories have different closure states

F1-009 shows the same problem through three contemporaneous trajectories.

AAE/UCS entered the utility docket while generation and transmission authorization remained open: FP-O.

Angelle Bradford Rosenberg testified after a settlement and approval motion had substantially narrowed the decision space, but before the Commission’s final vote: FP-C.

Dorothy Lively’s later construction-stage complaint occurred after original siting was no longer realistically open, yet traffic enforcement and construction-impact mitigation remained live: FP-O relative to that target.

The important object is therefore not “how far along was Hyperion?” but “how far along was the decision trajectory the act was trying to affect?”

4.6.3 Operation does not automatically imply foreclosure

F1-047 extends the point into the operating stage. The Fairwater facility was already operating when plaintiffs filed a nuisance complaint. The original siting and construction choice was not open, but the alleged noise was ongoing and potentially abatable. Damages, mitigation, fan operation, and injunctive relief remained materially changeable. The act was therefore coded FP-C, not mechanically FP-F.

The empirical rule that follows is:

FP is referent-specific. This is one of the clearest places where the Grammar prevents an apparently convenient but substantively wrong compression.

4.7 Actor Dependence (AD): standing and lived evidence are not authority

Actor Dependence asks what is lost when the acting subject is substituted. The confirmatory cases show why this cannot be inferred from formal power, channel, or rhetorical intensity.

4.7.1 Hyperion: low- and high-dependence acts in the same project

F1-009 contains a particularly useful within-project contrast.

The AAE/UCS regulatory motion is coded AD-L. Its legal and public-interest claims about utility arrangements, transparency, customer risk, and process can largely be reproduced by another similarly situated intervenor.

Rosenberg’s public testimony is also conservatively coded AD-L for its principal claims. Although she speaks as a member of the public and includes AI-specific concerns, her principal arguments about water demand, rates, jobs, oversight, and precedent are transferable claims rather than irreducibly first-person exposure evidence; they do not depend on her unique identity for their evidentiary content.

Dorothy Lively’s complaint is different. She lives at a principal construction entrance and reports trucks or workers using or parking on her property, dust or sand intrusion, difficulty leaving her driveway, and other location-specific construction effects. Substituting a generic resident would remove part of the evidence itself. The act is therefore AD-H.

All three are broadly contestatory acts associated with the same project. The difference lies not in how forcefully they oppose or condition realization, but in whether the actor is substantively replaceable.

4.7.2 Fairwater: high AD inside a formal legal channel

F1-047 blocks another shortcut. The plaintiffs’ federal nuisance complaint is highly formal and enters a court, but its Actor Dependence is AD-H, not low. The complaint contains direct sleep, work-shift, home-use, and property-specific noise evidence. A different plaintiff could reproduce the legal theory, but not the same lived and property evidence.

Formalization therefore does not make an act actor-independent.

4.7.3 AD is not a power scale

The converse is F1-044. Hamby WSC has actual constitutive authority over the water infrastructure it authorizes, yet the act is coded AD-M, not automatically high. AD does not ask how much power the subject has. It asks how much relevant act content is lost under substitution.

The distinction can be stated compactly:

This matters later in Section 6, where authority becomes one institutional feature of the path from consequence to realization. Treating AD as authority here would make that later test circular.

4.8 What the executed coding adds beyond visible Mode

Frame One contains repeated instances in which visible form remains similar while the executed RO/FP/AD assignments differ: petitions can be conditional or oppositional; one project can contain open, constrained, and foreclosed referents; and contestatory acts can carry low or high Actor Dependence. These observations show why the proposed Grammar is worth testing.

They do not establish that the Grammar is a reproducible measurement instrument. The same coder executed the primary classification and the matching reliability archive is incomplete. The defensible inference is therefore narrower:

Visible Participation Mode, project stage, and ordinary form descriptors do not reproduce all distinctions represented in the executed RO/FP/AD coding. Whether independent coders recover the same distinctions remains untested.

This is hypothesis-generating discrimination evidence, not validation.

4.9 Proposition A status

PropositionA_Discrimination = ObservedInExecutedCoding_HypothesisGenerating

PropositionA_Reliability = NotReconstructable_ArchivalLoss

The proposition survives as a testable analytical claim because the executed coding contains non-fungible distinctions. Its empirical validation is deferred to a prospectively archived paired-coder replication.

5. From Participation Taxonomy to Consequence

The consequence classifications use the frozen Packet-B rules in Appendix D; case evidence and source registers are in Supplement S1. The question here is not whether participation is generically “effective,” but whether observed acts and configurations stop at different consequence levels.

Because Manifestness and Materiality were not prospectively frozen, the strongest executed comparison uses Participation Mode and RO/FP/AD, with the secondary 3M readings used only to organize examples.

5.1 Analysis population and counting rule

At the August 28, 2026 cutoff, 14 of 21 confirmatory cases had mature twelve-month windows:

Participatory Consequences are defined relative to an observed act or act configuration, so the consequence inventory uses the 11 mature Present cases. F1-012 and F1-020 are SearchedNotFound, not zero-consequence cases; F1-001 is Unknown and not assessable. Consequence classes are non-exclusive and must not be summed into a score.

5.2 Descriptive consequence inventory

The table is a descriptive inventory of the executed mature case set, not an estimate of prevalence, relative effectiveness, or transition probability.

Table 3. Descriptive inventory of mature Participatory Consequences in Frame One

Consequence class	Present	Denominator	Interpretation
Informational	8	11	claims, evidence, or concerns enter or alter the recipient’s information state
Procedural	5	11	participation enters or produces a distinct review, hearing, adjudication, ballot, or authorization step
InstitutionalReturnPath	3	11	a durable complaint, reporting, review, clarification, or future-participation route is created or altered
SubjectReadiness	0	11	all 11 remain Unresolved
Distributional	1	11	allocation of benefits, burdens, compensation, or comparable terms changes
RealizationLevel	3	11	participation is supportably associated with conditioning or enabling of realization
NullNoDetectedChange	1	11	qualifying act observed, complete endpoint search, and no other consequence class detected

The main descriptive asymmetry is:

Information reaching an institution and realization changing are therefore empirically different events.

5.3 M5 does not encode consequence

M5 contestatory/claim-making participation provides the cleanest within-Mode comparison. The observed variation is informative, but the sample is too sparse and heterogeneous to identify why some configurations reach realization-level consequences and others do not.

The comparison below asks only which candidate conditioning differences are visible across representative M5 cases—route, revisability, authority relation, and documented outcome. It does not identify which feature drives conversion.

Table 4. Descriptive M5 comparison — candidate conditioning differences, not causal estimates

Case	Focal/operative route	Referent revisability	Participant authority	Observed consequence endpoint
F1-016 Richmond County	informal stop-project petition; no focal formal entry	substantially foreclosed/remedial	none	bounded Null
F1-017 Harwood	municipal hearing route	open	none	informational only
F1-042 Project Matador	protest plus later environmental process	open in later statutory route	none	informational + procedural; no attributed realization change
F1-025 Lancaster	multi-act municipal/zoning/CBA configuration	varies F→C→O by referent	none	realization conditioning, L2 configuration linkage
F1-048 Port Washington	municipal annexation/TID process plus later direct-legislation route	open	none	realization conditioning, L3 institutional attribution

The table exposes candidate conditioning differences—route, revisability, act density, institutional authority, legal setting, bargaining opportunity, timing—but does not isolate any one of them. Port Washington may differ because of institutional configuration, because its documentary record is stronger, or because of both; Frame One cannot distinguish those explanations.

The empirical result is therefore a negative structural finding:

M5 does not encode consequence. No claim is made about which M5 configuration is more effective or which institutional condition causes conversion.

5.4 Case-level consequence matrix

Table 5. Descriptive consequence-class matrix for the 11 mature Present cases

The consequence classes are non-exclusive: a case may exhibit several simultaneously. The counts are not additive scores and must not be read as a participation-success scale.

Case	Informational	Procedural	Return path	Distributional	Realization	Null	Strongest linkage
F1-008 Jeffersonville	P	N	N	U	N	—	L2
F1-009 Hyperion	P	P	P	U	U	—	L3
F1-013 Madison County	P	N	N	N	N	—	L3
F1-016 Richmond County	N	N	N	N	N	P	L0
F1-017 Harwood	P	N	N	N	N	—	L2
F1-025 Lancaster	P	P	P	P	P	—	L2
F1-027 Aiken County	U	N	N	N	N	U	L0
F1-032 Abilene	P	N	N	N	N	—	L3
F1-042 Project Matador	P	P	N	N	N	—	L3
F1-044 Frontier	N	P	N	U	P	—	L3
F1-048 Port Washington	P	P	P	N	P	—	L3

P = Present; N = NoDetectedChange; U = Unresolved. SubjectReadiness is omitted because all 11 are U.

Three boundaries are visible. First, one positive consequence does not force another: Abilene reaches information without attributed realization change, while Project Matador reaches information and procedure without realization change. Second, Unresolved is not converted into NoDetectedChange; Hyperion remains unresolved at realization level despite independent project conditioning. Third, Null requires a documented act plus a complete negative consequence search; project continuation alone is insufficient.

5.5 Where the transition stops

The mature records illustrate three distinct stopping points.

Act → Null. F1-016’s stop-project petition qualifies as participation but produces no detected informational, procedural, institutional, distributional, or realization consequence inside the frozen window. Later statutory participation cannot be back-propagated into that earlier act.

Act → information, not realization. F1-008, F1-013, F1-017, and F1-032 all reach informational consequence without attributed realization change. F1-032 is the clearest L3 example: the City Manager directly answers public concerns, but no focal decision or realization change is attributed to the comments.

Act → procedure, not requested realization change. F1-042 enters formal environmental adjudication, receives reasoned institutional disposition, and still produces no attributed realization change. Hyperion likewise has substantial informational/procedural consequences and a return path while focal realization attribution remains unresolved.

These cases jointly reject both shortcuts: “the project proceeded, therefore participation did nothing” and “officials responded, therefore participation changed realization.”

5.6 Realization consequence without project termination

Three mature cases are RealizationLevel=Present, and none is a focal-project denial, withdrawal, relocation, or reduction.

• F1-025 Lancaster — L2 configuration linkage: the project proceeds under binding environmental, operational, reporting, community-engagement, local-benefit, and financial commitments; the consequence is conditioning without stop.
• F1-044 Frontier — L3 direct-authority enabling: Hamby Water Supply Corporation authorizes project-specific water infrastructure that is subsequently implemented; the consequence is enabling rather than resistance.
• F1-048 Port Washington — L3 non-veto association sequence: the City later attributes zoning and sustainability conditioning to community input; a separate direct-legislation route also creates a return path. Counterfactual effect remains unidentified.

The realization class therefore captures conditioning or enabling, not merely termination.

5.7 Readiness remains measurement-limited

SubjectReadiness is Present in 0/11 and Unresolved in 11/11. Public archives show what people said, filed, petitioned, litigated, or authorized far better than whether participation made them better informed, more capable, more organized, or more likely to re-enter later.

ReadinessObservability = MeasurementLimited

This is a measurement gap, not evidence of a zero readiness effect.

5.8 Linkage strength

Across the 11 mature consequence-assessable cases, strongest case-level linkage is L3 in 6 cases, L2 in 3, L0 in 2, and L4 in 0. L3 is direct documentary attribution, not sole causation. No mature case combines direct attribution with the independent corroboration required for L4.

Accordingly:

5.9 Empirical consequence summary

The mature records support three separations:

The first arrow is directly challenged within frozen M5. The second is challenged by bounded Null. The third is challenged by informational/procedural cases without realization conversion. Conversely, Lancaster, Frontier, and Port Washington show that realization consequence can occur while the focal project proceeds.

These are case-set distinctions, not a fitted transition law. Their value is to prevent the paper from collapsing voice, response, procedure, realization, and termination into one variable.

6. Responsive Floor Test: From Consequence to Realization

Section 5 establishes that:

Section 6 asks what distinguishes cases that cross the second boundary. The taxonomy makes the reason for this question explicit: Frame One does not treat heterogeneous participatory types as a common “participation treatment,” and Section 5 shows that even the same broad M5 form can end in different consequence profiles. The registered contrasts instead examine access, channel structure, route to binding authority, and revisability, while Contrast E asks whether project realization differs descriptively between Present and SearchedNotFound.

6.1 Substantive uptake and realization are not the same variable

The A–D contrast work uses SubstantiveUptakeObserved, which is broader than Packet-B RealizationLevel=Present.

Uptake-positive mature configurations are: – F1-009 Hyperion; – F1-025 Lancaster; – F1-044 Frontier; – F1-048 Port Washington.

Packet-B realization-positive cases are: – F1-025; – F1-044; – F1-048.

Hyperion remains uptake-positive at configuration level while realization attribution is unresolved. The variables are not harmonized after outcome inspection.

6.2 Institutional contrasts

Table 6. Descriptive institutional contrasts in the mature case set

Contrast	Mature comparison	Observed result	Status
A — required procedural access	present  vs absent/not established	procedure entry 9/9 vs 0/2; uptake 4Y/4N/1U vs 0Y/2N	enabling for entry; substantive outcome unresolved
C — route to binding authority	clear route ; no route ; unresolved/potential	clear-route cases include both uptake and non-uptake	pathway-compatible; comparator sparse
D — revisability	open ; foreclosed ; mixed	all four uptake-positive configurations contain an open/later-open trajectory, but many open trajectories do not uptake	enabling, not sufficient

Contrast A strongly differentiates procedure entry, partly because access and entry are structurally related; it does not determine substantive uptake. Contrast C shows the same pattern: a live route to a binding authority is a structural opportunity, not an outcome. Contrast D likewise shows that open or reopened alternatives are present in every uptake-positive configuration, while many open configurations do not uptake. The small negative cells prevent stronger inference, but the pattern still justifies preserving entry, substantive uptake, and realization as separate outcome objects rather than collapsing them into a single participation-success variable.

F1-025 remains the clearest within-project revisability sequence. The primary stop-project act addresses an already committed phase (FP-F), a later zoning appeal a constrained but reviewable path (FP-C), and later phase-two comment an open decision (FP-O). F1-025-A02 remains in the packet but shares the first-phase FP-F state and therefore does not add a new step to the compact trajectory.

Table 7. Contrast B operability

Pre-specified comparison	Required comparator	Frame One result	Adjudication
target-controlled vs sufficiently independent/external channel	clean mature target-controlled primary-act cell	comparator absent	Pre-specified contrast not exercised — comparator cell absent.

6.3 Contrast E: precision gate unmet

Contrast E is the outcome-independent comparison closest to a domain-level Responsive Floor test. At the cutoff it contains 11 mature Present cases, two mature SearchedNotFound cases, and one mature Unknown. The frozen inferential rule requires at least 20 determinate projects in each binary cell plus the Wilson half-width threshold.

With a silent cell of , the contrast is not inferentially exercised. The visible project-outcome counts are retained only as a descriptive inventory in Appendix K.

ContrastE = DescriptiveOnly_PrecisionGateNotMet

No Fisher test, post-hoc category collapse, or favorable binary is promoted into the proposition adjudication.

6.4 Association-sequence terminology and causal standard

The manuscript uses documented association sequence for a traceable institutional-record sequence linking participation, an institutional process, and a later consequence or realization state. This is a documentary claim about ordering and attribution.

It does not mean causal identification, independent effect, necessity, sufficiency, or general mechanism. Mechanism-consistent is reserved for sequences compatible with a participatory causal account while material rivals remain. Identified mechanism would require substantially stronger counterfactual evidence; none is claimed here.

L3 institutional self-attribution can therefore support documentary linkage while remaining vulnerable to strategic self-presentation, pre-existing plans, and other actors’ influence. Even L4 would be necessary but not sufficient for mechanism identification without a design that materially reduces principal rival explanations.

6.5 Realization-level cases and competing explanations

Table 8. Mature realization-consequence cases: documentary linkage, not causal effect

Case	Authority position / path	Competing determinants visible in the record	Linkage	Defensible inference
F1-025 Lancaster	subjects lack binding authority; residents/advocates → municipal/developer process → CBA / prospective governance	prior commitments; zoning law; developer strategy; bargaining; electricity/infrastructure constraints; fiscal/local-benefit incentives	L2	mechanism-consistent mediated pathway; contribution not isolated
F1-044 Frontier	participating institution itself holds authority; water authorization → implementation	engineering/easement constraints; project demand; developer/infrastructure schedule	L3	direct-authority enabling; comparator, not non-veto evidence
F1-048 Port Washington	residents lack binding authority; community input → City decision environment → zoning/sustainability conditioning; separate direct-legislation return path	municipal legal authority; annexation/TID structure; developer strategy; fiscal bargaining; infrastructure; timing; other actors	L3	documented L3 non-veto association sequence in the institutional record; counterfactual effect unidentified

Port Washington supplies the strongest non-veto sequence:

The City directly states that community input informed zoning and sustainability commitments. That is stronger than temporal sequence, but it remains institutional self-attribution: the record does not establish that the same conditions would have been absent without participation. Lancaster is weaker at L2.

In both cases, municipal legal authority, developer strategy, infrastructure and fiscal conditions, bargaining, timing, and other actors remain plausible co-determinants; the record documents a participatory association sequence rather than isolating participation’s independent effect.

A near-comparator among Port Washington, Abilene, and Harwood holds several observed features approximately constant—public participation, no participant binding authority, institutional channel, recipient authority, and open primary FP—while attributed realization differs. The cases remain dissimilar in law, project stage, bargaining opportunity, sponsor strategy, act density, and infrastructure; the comparison is diagnostic, not causal.

The negative cases are equally important. F1-016 records a true bounded Null; F1-017 and F1-032 reach response without realization change; F1-042 reaches formal adjudication without attributed realization conversion; F1-013’s proposed local-benefit condition is not adopted; F1-009 reaches strong informational/procedural consequences while realization attribution remains unresolved. These cases prevent access, response, or conditioning elsewhere from being redefined as automatic participation success.

6.6 F1-044 identity sensitivity

The full 49-row identity audit confirms one cross-row physical-campus collision: pilot F1-043 and confirmatory F1-044 refer to the same Frontier/Stargate Shackelford campus. No additional collision was identified among the other 1,175 pairs after mechanical screening and external alias checks.

F1-044 is retained in the primary tables because deleting a sampled sponsor/project row after randomization would change the frozen sample; F1-043 is pilot-only, so there is no confirmatory double count. The leave-one-out sensitivity reduces mature Present 11→10, Procedural Present 5→4, RealizationLevel Present 3→2, and L3 cases 6→5. It removes the direct-authority comparator and sole primary-signature M8/RO-S case, but it does not remove the Port Washington L3 non-veto association sequence, the Lancaster configuration, transition separation, or the failure of the Contrast-E precision gate.

6.7 Responsive Floor Test result and Proposition C

Two of the three mature realization-positive cases involve subjects without binding authority; one is direct authority. This is sufficient to reject only the narrow descriptive claim that every realization-relevant participatory record in Frame One reduces to the participating subject exercising binding authority itself. It is not sufficient to estimate a non-veto effect.

The Responsive Floor Test therefore returns a split result:

ResponsiveFloorTest_AssociationSequence = DocumentedAssociationSequenceObserved

ResponsiveFloorTest_Aggregate = NotAdjudicated

The association-sequence result rests principally on Port Washington’s L3 institutional self-attribution and is best read as a documented non-veto association sequence requiring causal follow-up. Lancaster supplies weaker L2 configuration evidence. Neither case identifies necessity, sufficiency, an independent effect, or a general mechanism.

The aggregate/domain burden remains unadjudicated because the silent cell is tiny, major realization conditions are heterogeneous, several outcomes are configuration-level, and seven cases remain right-censored.

ResponsiveFloor_FrameOne = NotAdjudicated.

Frame One contains a documented institutional-record association sequence linking non-veto participation to later realization conditioning. The counterfactual contribution of participation is unidentified, and the outcome-independent cross-case evidence is too sparse and heterogeneous to adjudicate the data-center domain as a Responsive Floor.

7. Time, Sentiment, and Institutionalization

The temporal extension tested a plausible 2026 hypothesis: focal resistance might become more frequent as AI capability and public concern increased. Appendix G preserves the specifications; the body reports only the result relevant to the paper’s claims.

7.1 Focal resistance frequency

Table 9. Standardized focal resistance-event rate by calendar period

Period	Focal broad-resistance episodes	Exposed project-month equivalents	Events per 100 project-months
pre-2026	23	152.18	15.11
2026 through Aug. 28	10	88.61	11.29

The crude 2026/pre-2026 rate ratio is approximately  with a wide approximate 95% interval of 0.36–1.57; age-adjusted exploratory models likewise fail to identify a positive calendar-2026 effect. The reader-facing conclusion is therefore:

The temporal data are insufficient to identify a positive calendar-2026 resistance-frequency effect; uncertainty is too wide to establish either an increase or its absence.

The frozen machine-readable adjudication remains Calendar2026ResistanceFrequencyIncrease = NotSupported, where NotSupported means that the positive hypothesis was not supported—not that a zero or negative effect was established. For reader-facing reporting, the equivalent label is PositiveEffectNotIdentified. F1-013 and F1-044 leave-one-out sensitivities do not change that interpretation.

7.2 Project age and institutionalization

Formal entry is not uniquely a late-stage phenomenon: many early acts already enter councils, commissions, courts, legislatures, or regulatory proceedings. Exploratory downstream-adaptation sensitivities point in a positive age direction but remain imprecise and coding-sensitive.

InstitutionalizationMaturation = NotIdentified_TemporalDataInsufficient.

The observed cases may motivate a future hypothesis that durable conversion into rules, conditions, mitigation, benefits, or return paths emerges later than initial participation. Frame One does not identify project age as the cause or establish a transition law.

7.3 Capability fear and contextual sentiment

The focal records predominantly concern electricity, water, pollution, noise, land/community character, subsidies, jobs/local benefit, transparency, process speed, and local control. No coded focal event establishes increasingly capable AI as the principal expressed cause of resistance.

AICapabilityEscalationFearMechanism = NotSupportedInFrameOne.

This does not imply that broader AI anxiety is absent. Survey sentiment, news framing, social discourse, and observed participation remain separate evidence classes; the news/social protocols were not completed as quantitative result corpora.

8. Formal Placement: Specification for Future Research

This section is a specification constrained by empirical contact, not an estimated model. For act , let

Frame One supports only a typed partial-transition representation:

The arrows denote observed ordering or documentary linkage where supported, not transition probabilities, causal effects, or coefficients. Richmond County stops at a bounded Null; Harwood/Abilene-type cases stop at informational consequence without attributed realization conversion; Port Washington reaches a documented institutional association sequence with counterfactual contribution unidentified. The within-M5 result imposes the central negative constraint:

A non-veto sequence may pass from participants lacking binding authority to an authoritative recipient without implying that the participants’ causal effect has been identified. Frame One therefore does not identify a universal Participation scalar, additive coefficient, universal Manifestness/Materiality/Mode effect, or direct generic Participation argument in the inherited realization mapping. Appendix A retains the fuller formal catalog and the conditions required for future estimation.

9. Evidence Status After Empirical Contact

9.1 Evidence-status ledger

Table 10. Proposition and result status

Claim / task	Present evidentiary status
3M Participation Taxonomy	proposed integrated taxonomy; Modality prospectively exercised; Manifestness/Materiality secondary descriptive; full validation not exercised
M5 contingency	structural negative finding: Mode does not encode consequence; explanatory comparison too sparse
Proposition A — Conditional Non-Fungibility	discrimination observed in executed single-coder coding; hypothesis-generating; independent reproducibility untested
Proposition B — Role Plasticity	UntestedInFrameOne; automation antecedent absent
Responsive Floor Test / Proposition C	Port Washington L3 documented association sequence; aggregate/domain NotAdjudicated; counterfactual contribution unidentified
Act / Consequence / Realization separation	directly illustrated in mature records
Participatory Readiness	0/11 Present; 11/11 Unresolved; public archives poorly suited to measurement
2026 frequency hypothesis	positive calendar-2026 effect not identified; wide uncertainty prevents an absence claim
capability-escalation fear	not identified as focal explanation in Frame One
Participation scalar / universal type effect	not warranted / not identified

Three empirical statements survive intact:

1. Mode does not encode consequence. The same frozen M5 Mode spans bounded Null, intermediate consequence, and realization conditioning.
2. The Grammar is worth replicating, not yet validated. In the executed single-coder, outcome-blind coding, RO/FP/AD preserve distinctions; the observation is hypothesis-generating and independent reproducibility is untested.
3. Non-veto participation appears in a documented institutional association sequence. Port Washington supplies direct institutional attribution, while causal contribution and domain-level responsiveness remain unidentified.

The title therefore names three linked analytical objects with deliberately asymmetric status:

3M Taxonomy: proposed + partly exercised
Relational Grammar: hypothesis-generating discrimination observed in executed single-coder coding; reliability unresolved
Responsive Floor Test: association sequence observed; domain not adjudicated

10. Limitations and Conclusion

The taxonomy makes several limitations more visible rather than less. More cases would help some problems but would not repair archival provenance, balance an observation strategy toward implicit participation, instantiate an absent automation antecedent, or turn documentary attribution into counterfactual identification.

10.1 Frame, unit, and taxonomic coverage limits

Frame One is a bounded, partly reconstructed research frame, not a probability sample of every U.S. data center, every AI infrastructure project, or societal participation generally. Eleven F1 mappings were directly recovered and 38 mechanically reconstructed; the original written ordering instruction is missing even though the reconstruction rule reproduces every recovered anchor. The randomized unit is a sponsor/project-announcement row, which need not remain identical to a unique physical-campus object as projects, landlords, tenants, branding, and phases evolve.

The full 49-row / 1,176-pair identity audit confirms one cross-row physical-campus collision: F1-043/F1-044. The other flagged Stargate-brand pairs are separate sites. Because F1-043 is pilot-only and F1-044 confirmatory, the known campus is represented once in confirmatory estimates. The existing F1-044 leave-one-out sensitivity preserves the headline conclusions while removing the sole primary M8/RO-S observation and the direct-authority comparator.

AI relevance is also heterogeneous in provenance. Twenty of 21 confirmatory cases explicitly pass a project-side AI gate; F1-013 passes on Moderate-confidence contextual/public linkage, and no case relies on governance-side AI relevance alone. Excluding F1-013 leaves the headline findings and negative temporal result unchanged. This does not eliminate the broader limitation that publicly AI-labeled infrastructure may differ systematically from less visibly AI-linked infrastructure.

Taxonomic coverage is uneven. Only M1, M4, M5, and M8 appear among the 16 frozen primary Modes. The secondary descriptive 3M pass yields 16/16 Explicit acts and Materiality of 10 Intangible, 5 Mixed, and 1 Tangible. Those numbers describe what this public-record design recovered, not what participation generally looks like.

The unobserved primary Modes also have identifiable observation requirements:

Unobserved Mode	Why Frame One is poorly positioned to recover it
M2 Evaluative / Feedback	structured feedback may sit outside project permitting and public-hearing records
M3 Behavioral / Choice	requires adoption, refusal, exit, purchase, or comparable behavioral traces
M6 Productive / Co-creative	requires contribution, production, repository, dataset, or collaborative-work records
M7 Adaptive / Reorganizational	requires longitudinal observation of training, workflow, role, or institutional reorganization

Their absence from primary-signature occupancy is therefore an observation-opportunity gap, not evidence of theoretical invalidity or substantive absence.

10.2 Observation and follow-up limits

Public archives expose formal hearings, petitions, dockets, official responses, lawsuits, protests, and many public complaints more readily than private negotiation, deleted platform activity, informal conversations, everyday adoption/non-adoption, unrecorded contribution, organizational adaptation, or antecedent readiness. SearchedNotFound therefore means a completed standardized public-record search found no qualifying focal act; it does not mean no human participation occurred. Unknown remains separate.

This asymmetry is now visible inside the taxonomy itself. Where a study lacks a credible channel for observing a participation type, non-observation should be represented as UnderObserved or Unresolved, not Absent. Frame One’s 16/16 Explicit primary-act occupancy is therefore a measurement result about observation opportunity as much as a descriptive result about the recovered cases.

Seven confirmatory projects are right-censored at the August 28 cutoff. They remain in the sample but are not assigned final Packet-B endpoints early. Post-window evidence can contextualize a frozen-window fact but cannot rewrite it.

10.3 Reliability and attribution limits

The registered inter-coder reliability statistic is not reconstructable. The archive preserves one complete blind return but not the complete matching pre-adjudication pair with sufficient provenance. The executed RO/FP/AD contrasts should therefore be read as single-coder outcome-blind discrimination evidence, not as evidence of reproducible classification. The study makes no reliability pass, failure, zero-reliability, or instrument-validation claim.

Manifestness and Materiality have an additional design limitation: they were not prospectively frozen Packet-A variables. Their present application is secondary descriptive coding from preserved acts and cannot validate the full 3M Taxonomy. A taxonomy-complete replication must freeze all three 3M dimensions and preserve paired raw pre-adjudication assignments.

Attribution is a separate limitation. L3 means direct documentary attribution, not causal necessity or sufficiency; no mature case reaches L4. Port Washington is notable because the City directly states that community input informed zoning and sustainability commitments. That statement may still coexist with municipal self-presentation, pre-existing policy plans, developer strategy, fiscal bargaining, infrastructure constraints, timing, and other actors’ influence. Lancaster is weaker at L2 configuration linkage.

The study can therefore document an institutional-record association sequence while leaving the counterfactual causal contribution unidentified.

10.4 Temporal and contextual limits

The temporal models are exploratory and sparse. They do not identify a positive calendar-2026 frequency effect, but the intervals are too wide to establish either an increase or its absence. The temporal data are also insufficient to infer a maturation effect, although downstream institutional responses are descriptively visible. Surveys, news, social discourse, and focal acts remain separate evidence classes. Broader AI anxiety may matter contextually, but this study does not infer a project-level mechanism from external sentiment alone.

10.5 What a stronger 3M Taxonomy and Responsive Floor Test require

A stronger empirical program should preserve the existing outcome-blind architecture while changing the observation strategy by participation type.

• Full 3M reliability: freeze Manifestness, Participation Mode, Materiality, RO, FP, and AD prospectively in one immutable act-level instrument; preserve complete paired pre-adjudication coding archives.
• M3 Behavioral / Choice: use consented longitudinal panels, documented adoption/non-adoption, procurement or usage traces, and exit/refusal records. Behavioral traces alone do not establish implicit participation; intentional relation to an identifiable realization object must be independently supportable.
• M6 Productive / Co-creative: observe repositories, contribution histories, dataset/content/code/research production, collaborative records, or other traceable contribution objects.
• M7 Adaptive / Reorganizational: observe longitudinal training, workflow change, role redesign, internal procedures, organizational records, interviews, or participant diaries.
• Implicit participation generally: use prospective repeated interviews, diaries, panels, or consented behavioral/organizational traces capable of connecting intentional conduct to a realization object even where the actor does not publicly label the conduct as participation.
• Observation opportunity: record, before outcome analysis, whether each study has a credible channel for each target participation type. No credible channel means UnderObserved/Unresolved, not Absent.
• Cross-channel validation: in at least a subsample, compare public records, participant reports, and behavioral/organizational traces to estimate what each source class systematically misses.
• Responsive Floor adjudication: retain outcome-independent sampling, include enough comparable Present and completed-search cases to satisfy the frozen precision rule, and measure major legal, infrastructure, commercial, and authority conditions strongly enough to reduce rival explanations.
• Readiness: measure subject knowledge/capacity before and after participation with prospective subject-level instruments rather than public archives alone.
• Concrete implicit-participation test in the data-center domain: a landowner who intentionally refuses a project easement or parcel transaction without making a public claim could be coded, if intent and project relation are independently documented, as Implicit · M3 Behavioral/Choice · Tangible. The refusal would not be inferred from silence or non-sale alone; evidence of intentional realization relation is required.
• Role Plasticity proving ground: agentic software development is a stronger domain than infrastructure for Proposition B because task transfer and human-role change can be observed longitudinally. Coding-agent telemetry can distinguish planning, execution, review, approval, override, and integration. Anthropic’s 2026 analysis of roughly 400,000 Claude Code sessions found a division in which users made most planning decisions while the agent made most execution decisions, providing an observable transition from human implementation toward specification, review, orchestration, and governance. A prospective study could freeze task units and test whether human participation shifts rather than disappears as execution becomes more automated.

This is not a request simply for a larger version of Frame One. It is a request for a mode-sensitive observation design. The 3M Taxonomy becomes methodologically useful when it tells the researcher not only how to classify an observed act, but also which acts the chosen evidence architecture is structurally unlikely to see.

10.6 Broader Gradual AGI implication

A natural question is whether the Gradual AGI framing is necessary if many focal concerns look like ordinary infrastructure politics. The empirical answer is precisely that realization is ordinary in this way. Increasing AI capability does not materialize directly into societal effect. It arrives through electricity systems, land, water, air permits, noise, fiscal distribution, zoning, local institutions, work practices, adoption choices, contribution, and recourse. Human participation enters those realization processes in multiple forms—sometimes explicitly, sometimes implicitly; sometimes intangibly through information and claim-making, sometimes tangibly through behavior, production, institutional action, or material implementation.

The Floor is therefore not simply where the Ceiling arrives. But neither is the Floor demonstrated here to be uniformly responsive to participation.

An AI-specific version of that claim would require a comparator showing that otherwise comparable AI-labeled and generic infrastructure systematically differ in participatory configuration or conversion. Frame One contains no generic-infrastructure comparator, so it cannot separate AI-specific participation dynamics from infrastructure-general ones.

Three results can coexist:

and:

The first is the taxonomic result: knowing the type does not settle what follows. The second is the non-veto association sequence result: the institutional record contains a sequence ending in a realization condition not reducible to the participant possessing formal authority, while causal contribution remains unidentified. The third is the conversion result: voice, procedure, response, or controversy cannot be treated as realization change merely because it occurs.

The data-center domain does not establish a Responsive Floor, validate the full 3M Taxonomy, or establish a reproducible Grammar instrument. It establishes a disciplined empirical program for Participation: classify the type, preserve the relational signature, observe the consequence, document the sequence, and refuse to infer any missing causal arrow.

References

Scholarly, series, and methodological references

Arnstein, Sherry R. 1969. “A Ladder of Citizen Participation.” Journal of the American Institute of Planners 35(4): 216–224. https://doi.org/10.1080/01944366908977225

Birhane, Abeba, William Samuel Isaac, Vinodkumar Prabhakaran, Mark Díaz, Madeleine Clare Elish, Iason Gabriel, & Shakir Mohamed. 2022. “Frameworks and Challenges to Participatory AI.” Proceedings of the Second ACM Conference on Equity and Access in Algorithms, Mechanisms, and Optimization (EAAMO ’22).

Delgado, Fernando, Stephen Yang, Michael Madaio, & Qian Yang. 2023. “The Participatory Turn in AI Design: Theoretical Foundations and the Current State of Practice.” arXiv:2310.00907.

Ekman, Joakim, & Erik Amnå. 2012. “Political Participation and Civic Engagement: Towards a New Typology.” Human Affairs 22(3): 283–300. https://doi.org/10.2478/s13374-012-0024-1

Fung, Archon. 2006. “Varieties of Participation in Complex Governance.” Public Administration Review 66(s1): 66–75. https://doi.org/10.1111/j.1540-6210.2006.00667.x

Marres, Noortje. 2015. Material Participation: Technology, the Environment and Everyday Publics. London: Palgrave Macmillan. https://doi.org/10.1007/978-1-137-48074-3

Nesta. 2021. Participatory AI for Humanitarian Innovation: A Briefing Paper. The framework was later catalogued by OECD.AI as the Participatory AI Framework.

OECD. 2026. Artificial Intelligence and the Future of Citizen Participation: Typology of Applications, Opportunities and Challenges for Democratic Innovation. OECD Public Governance Reviews. https://doi.org/10.1787/a1ee2e0a-en

Pateman, Carole. 1970. Participation and Democratic Theory. Cambridge: Cambridge University Press. https://doi.org/10.1017/CBO9780511720444

Rowe, Gene, & Lynn J. Frewer. 2005. “A Typology of Public Engagement Mechanisms.” Science, Technology, & Human Values 30(2): 251–290. https://doi.org/10.1177/0162243904271724

Rydin, Yvonne, & Lucy Natarajan. 2016. “The Materiality of Public Participation: The Case of Community Consultation on Spatial Planning for North Northamptonshire, England.” Local Environment 21(10): 1243–1251. https://doi.org/10.1080/13549839.2015.1095718

Theocharis, Yannis, & Jan W. van Deth. 2018. “The Continuous Expansion of Citizen Participation: A New Taxonomy.” European Political Science Review 10(1): 139–163. First published online November 21, 2016.

Villi, Mikko, & Janne Matikainen. 2016. “Participation in Social Media: Studying Explicit and Implicit Forms of Participation in Communicative Social Networks.” Media and Communication 4(4). https://doi.org/10.17645/mac.v4i4.578

Anthropic. 2026. “Agentic Coding and Persistent Returns to Expertise.” June 16, 2026. Privacy-preserving analysis of approximately 400,000 Claude Code sessions from October 2025 through April 2026. https://www.anthropic.com/research/claude-code-expertise

MIT Ad Hoc Committee on AI Use in Teaching, Learning, and Research Training. 2026. Final Report of the Ad Hoc Committee on AI Use in Teaching, Learning, and Research Training. Report dated August 13, 2026; publicly released August 25, 2026. Massachusetts Institute of Technology.

Krippendorff, Klaus. 2018. Content Analysis: An Introduction to Its Methodology. 4th ed. Thousand Oaks, CA: SAGE.

W.H.L. & GPT-5.5. 2026. “Gradual AGI as Synchronization for Transformative Adoption.” Champaign Magazine, June 29, 2026. https://champaignmagazine.com/2026/06/29/gradual-agi-as-synchronization-for-transformative-adoption/

W.H.L. & Claude (Sonnet 5). 2026. “Ceiling, Floor, and Slope: A Falsifiable Dynamical Model of Synchronization for Gradual AGI.” Champaign Magazine, July 4, 2026. https://champaignmagazine.com/2026/07/04/ceiling-floor-and-slope-a-falsifiable-dynamical-model-of-synchronization-for-gradual-agi/

W.H.L., Claude Sonnet 5, & GPT-5.5. 2026. “Gradual AGI as Optimization: A Conceptual Framework.” Champaign Magazine, July 22, 2026. https://champaignmagazine.com/2026/07/22/gradual-agi-as-optimization-a-conceptual-framework/

W.H.L. & Claude (Sonnet 5, Opus 5). 2026. “Gradual AGI as Optimization: Formal Models and Empirical Tests.” Champaign Magazine, July 24, 2026.

W.H.L. & Claude (Opus 5). 2026. “Gradual AGI as Contestation: A Framework for Governance.” Gradual AGI Series #7. Champaign Magazine, August 3, 2026. https://champaignmagazine.com/2026/08/03/gradual-agi-as-contestation-a-framework-for-governance/

W.H.L., Claude Opus 5, & GPT-5.6 Sol. 2026. “Gradual AGI as Contestation: Measuring Governance Under Empirical Contact.” Gradual AGI Series #8. Champaign Magazine, August 11, 2026. https://champaignmagazine.com/2026/08/11/gradual-agi-as-contestation-measuring-governance-under-empirical-contact/

W.H.L., GPT-5.6 Sol, & Claude Sonnet 5. 2026. “Gradual AGI as Participation: Subjects, Acts, and Consequences.” Gradual AGI Series #9. Champaign Magazine, August 27, 2026. https://champaignmagazine.com/2026/08/27/gradual-agi-as-participation-subjects-acts-and-consequences/

Contextual survey and discourse sources used in Section 7

Kennedy, Brian, Eileen Yam, Emma Kikuchi, Isabelle Pula, & Javier Fuentes. 2025. “How Americans View AI and Its Impact on People and Society.” Pew Research Center, September 17, 2025. Survey fielded June 9–15, 2025. https://www.pewresearch.org/science/2025/09/17/how-americans-view-ai-and-its-impact-on-people-and-society/

Quinnipiac University Poll. 2026. “The Age of Artificial Intelligence: Americans’ AI Use Increases While Views on It Sour.” March 30, 2026. Survey fielded March 19–23, 2026. https://poll.qu.edu/poll-release?releaseid=3955

Pew Research Center. 2026. “Americans and AI 2026: Chatbots, Smart Devices and Views on Impact.” June 17, 2026. Survey fielded February 17–23, 2026. https://www.pewresearch.org/internet/2026/06/17/americans-and-ai-2026-chatbots-smart-devices-and-views-on-impact/

Volcovici, Valerie, & Jason Lange. 2026. “Americans Wary of AI-Driven Data Center Boom, Reuters/Ipsos Poll Shows.” Reuters, June 11, 2026. https://www.reuters.com/world/us/americans-wary-ai-driven-data-center-boom-reutersipsos-poll-shows-2026-06-11/

The Soufan Center. 2026. “A Year of Monitoring: Anti-Data Center Narratives and Threats.” IntelBrief, July 21, 2026. https://thesoufancenter.org/intelbrief-2026-july-21/

Conrad, Jennifer. 2026. “Geoffrey Hinton’s Blunt Warning About Elon Musk and Mark Zuckerberg Got the Biggest Applause at Ai4.” Inc., August 5, 2026. https://www.inc.com/jennifer-conrad/geoffrey-hintons-blunt-warning-about-elon-musk-and-mark-zuckerberg-got-the-biggest-applause-at-ai4/91385482

Schmelzer, Ron. 2026. “Three AI Pioneers Clash Over Jobs, Regulation and the Future of AI.” Forbes, August 6, 2026. https://www.forbes.com/sites/ronschmelzer/2026/08/06/hinton-li-ng-ai-jobs-regulation/

Primary project evidence

Case-level factual claims are supported by the project-specific outcome-blind Packet A records, Packet B records, status registers, and source logs created for Frame One. The full 21-case evidence and primary-source register is published separately as Supplement S1 — Confirmatory Case Evidence and Primary-Source Register. This keeps the compact reference list from duplicating project-level docket and local-record entries while preserving the evidentiary backplane.

Appendices

Appendix A — Notation and Formal Objects

This appendix fixes the notation used in the formal extension. It does not introduce additional empirical claims.

A.1 Inherited realization notation

Symbol	Meaning	Status in #10
	capability or capability class	inherited
	time indices	inherited
	group / population position for realized outcome	inherited
	capability potential at time	inherited
	realized outcome for group  at time	inherited
	realization mapping	inherited and unchanged
	group-specific exposure state	inherited
	reach component of exposure	inherited
	susceptibility component of exposure	inherited
	recourse component of exposure	inherited
	governing arrangement relevant to capability  at time	inherited
	actor aggregation / effective gate object in CFS	inherited; participation-specific change not identified
	effective threshold, where used in CFS representations	inherited/admissible; participation-specific shift not identified

The inherited realization mapping remains:

with:

For compact notation only:

 is a structured state, not a scalar.

A.2 Participation-specific formal objects

Symbol	Meaning
	realization object / referent
	3M Participation Taxonomy type for act : Manifestness, Modality/Participation Mode, Materiality
	participatory signature for act : taxonomy + RO/FP/AD
	realization-specific Participatory Configuration observed over
	structured set/distribution of participatory signatures in configuration
	Participatory Consequence of class  associated with realization
	consequence class index
	linkage-support grade, L0–L4
	typed conversion pathway from consequence to a changed inherited realization state
	actors possessing binding authority over realization  at time
	Participatory Subjects represented in configuration
	Participatory Configuration whose subjects do not themselves hold binding authority over
	information state of actor , used only in the information-to-authority path
	procedural/decision-processing state of the governing arrangement

The primary empirical-contact representation is:

followed, where a realization-relevant state change is observed, by:

The two transitions are partial and can fail independently. In particular:

Manifestness and Materiality are taxonomy-level objects in the formal representation, but they are not treated as frozen confirmatory fields in Frame One. Any future model using them as predictors requires prospective coding and reliability evidence.

A.3 Consequence-class index

where:

•  — Informational;
•  — Procedural;
•  — Institutional / return-path;
•  — Subject-readiness;
•  — Distributional;
•  — Realization-level;
•  — Null / no detected change.

The classes are non-exclusive and unordered.

A.4 Non-veto pathway

A non-veto configuration satisfies:

A mediated path can then be represented as:

This notation identifies an authority-mediated path; it does not establish sole causation.

A.5 Formal non-claims

Frame One does not identify:

• a universal Participation scalar ;
• a stable function ;
• an additive participation coefficient ;
• a measured participation-specific ;
• a participation-specific threshold coefficient;
• a direct generic Participation argument inside .

Canonical formal statuses:

• NewParticipationScalar = NotWarranted
• DirectPiArgumentInMu = NotIdentified
• ParticipationToG = Observed
• ParticipationToRecourse = ObservedInSelectedCases
• ParticipationToPhi = AdmissibleButNotIdentified
• ParticipationToThreshold = AdmissibleButNotIdentified
• SubjectReadinessRecursion = StructurallySpecified_EmpiricallyUnresolved

A.6 Responsive Floor Test: adjudication criteria

The Responsive Floor Test has two analytically separate components.

Association-sequence component. Determine whether the record contains a traceable institutional sequence linking a pre-specified Participatory Configuration to a later binding or material realization condition. A mediated non-veto sequence may satisfy this component without identifying counterfactual causal contribution.

Aggregate/domain component. A domain-level Responsive Floor classification requires: 1. an outcome-independent frame and comparator; 2. determinate participation states; 3. adequate pre-specified precision; 4. sufficient comparability, stratification, or modeling of major realization conditions; 5. evidence that participatory configuration adds explanatory value; 6. separation of documentary linkage from counterfactual causation.

The allowed aggregate verdicts are Supported, NotSupported, or NotAdjudicated. NotAdjudicated is required when the comparison is too sparse, heterogeneous, censored, or otherwise inoperable to distinguish support from non-support at the pre-specified standard.

Repeated non-veto association sequences can strengthen the association-sequence component but do not substitute for the aggregate/domain burden.

A.7 From documented association sequence to identified mechanism

The manuscript uses documented association sequence for a record-linked sequence and mechanism-consistent when that sequence is compatible with a causal account but rival explanations remain. Identified mechanism is reserved for substantially stronger evidence.

For the present architecture, L4 is necessary but not sufficient. Independent corroboration must be joined by pre-specified temporal/process evidence and a design or comparison that materially reduces the principal competing explanations. Repeated L3 cases alone do not identify a mechanism.

Appendix B — Participation Taxonomy and Act-Level Codebook

This appendix is the reader-facing taxonomy and codebook for the act level. Manifestness, Modality, and Materiality define the 3M Participation Taxonomy; Modality takes one of eight Participation Modes; RO, FP, and AD remain the separate Relational Grammar. Internal identifiers M1–M8 are retained for reproducibility. Only the Packet-A Act Form field (interpreted here as Modality / Participation Mode) and RO/FP/AD were frozen confirmatory fields in Frame One; Manifestness and Materiality are taxonomy dimensions added for transparent descriptive interpretation and prospective replication.

B.1 Inclusion gate for a Participatory Act

A focal act must satisfy all of the following:

1. eligible human or human-organized Participatory Subject;
2. observable act;
3. intentional relation to AI/AI realization as defined by the project protocol;
4. identifiable realization object/referent;
5. sufficient evidence to assign the act to the focal case and observation window.

Affectedness alone is insufficient. Later outcome is never used to decide whether the earlier act qualifies.

B.2 Manifestness

Value	Rule
Explicit	the act overtly presents itself as addressing, supporting, opposing, conditioning, redirecting, reporting on, authorizing, or otherwise intervening in an identifiable realization
Implicit	the intentional relation to the realization is established through behavior, practice, contribution, adaptation, or organization without an overt participatory claim

Boundary: implicit participation still requires observable, attributable intentional action. Affectedness, passive exposure, latent preference, or an unexpressed attitude does not qualify.

Public-expression rule: an intentionally public or socially accessible expression of AI-related sentiment is a Participatory Act. Distrust, worry, praise, criticism, humor, parody, or mockery is normally M1 unless the operative act is structured feedback (M2), deliberation (M4), claim-making (M5), or another Mode.

Frame One status: descriptive taxonomy only; not a frozen Packet-A field and not used for retrospective confirmatory counts.

B.3 Modality and Participation Mode

Code	Reader-facing label	Operational description
M1	Expressive / Discursive	public expression, speech, statement, publication, commentary, joke, meme, parody, mockery, praise, criticism, or discourse not better classified below
M2	Evaluative / Feedback	evaluation, review, feedback, reporting, assessment, poll/survey response, or other structured response to an AI-related realization
M3	Behavioral / Choice	observable choice, adoption/non-adoption, exit, purchase/use decision, or other behavior bearing directly on realization
M4	Deliberative / Consultative	participation in consultation, deliberation, hearing, advisory process, or structured discussion
M5	Contestatory / Claim-making	petition, objection, protest, challenge, appeal, lawsuit, demand, or other claim-making act
M6	Productive / Co-creative	production, contribution, collaboration, dataset/content creation, or co-creation bearing on AI realization
M7	Adaptive / Reorganizational	training, institutional adaptation, role reorganization, workflow change, or capacity-building response
M8	Authoritative / Constitutive	act undertaken through binding or constitutive institutional authority, including approvals, authorizations, rule-making, or formal institutional acts that themselves alter realization conditions

Participation Mode describes what visibly happens. It does not determine orientation, revisability, or Actor Dependence.

B.3.1 Unit and precedence rule

Mode is assigned to the operative function of a specific act by a specific Participatory Subject.

The same distrust, enthusiasm, or concern can travel through different Modes: a public post is normally M1; a structured poll response M2; a boycott or adoption decision M3; consultation participation M4; a petition or protest M5; a productive contribution M6; organizational adaptation M7; and a binding institutional authorization or denial M8.

A broader episode is not forced into one Mode. It may contain multiple subjects and multiple acts:

Where one act performs more than one function, assign the primary Mode from the function without which the act would cease to be the act being coded. A secondary Mode may be retained only where the same act genuinely performs a second analytically distinct function; secondary coding may not be used to rescue ambiguity or inflate category occupancy.

B.4 Materiality

Value	Rule
Intangible	immediate act operates primarily through expression, information, evaluation, interpretation, symbolic positioning, or other communicative/knowledge-bearing action
Tangible	immediate act directly changes behavior, an artifact, resource allocation, organizational practice, legal/institutional state, or a material component of realization
Mixed	communicative/interpretive and behavioral/material/constitutive operation are jointly part of the act and cannot be cleanly separated

Materiality is about the act, not the consequence. An intangible act may produce a tangible consequence; a tangible act may produce no detected consequence.

Frame One status: descriptive taxonomy only; not a frozen Packet-A field and not used for retrospective confirmatory counts.

B.5 Realization Orientation (RO)

RO asks: what relation does this act take toward the specific realization referent?

Code	Label	Rule
RO-S	Supportive	supports, enables, advances, or seeks continuation of the referent
RO-C	Conditional / Redirective	seeks pause, condition, modification, redesign, alternative path, or contingent continuation
RO-O	Oppositional	seeks prevention, denial, termination, cancellation, or blocking of the referent
RO-R	Reporting / Nondirective	reports, documents, evaluates, or supplies information without a directive realization demand

RO is not general sentiment toward AI. A subject may support AI while opposing one facility, or oppose one configuration while supporting a modified version.

B.6 Foreclosure Position (FP)

FP asks: how open is the meaningful realization choice for the specific act–referent pair when the act occurs?

Code	Label	Rule
FP-O	Open	substantively different realization choices remain institutionally or practically available
FP-C	Constrained	meaningful alternatives remain, but the available set is materially narrowed by prior commitments, procedure, or project state
FP-F	Foreclosed / Remedial	the principal targeted choice is substantially settled; the act is remedial, retrospective, or directed at reopening an otherwise foreclosed path
FP-U	Unresolved	outcome-blind evidence is insufficient to determine the position

Boundary rule: project stage is not FP. One project can contain multiple simultaneous referents with different foreclosure positions.

B.7 Actor Dependence (AD)

AD asks: how much substantive evidentiary, experiential, or standing-related content would be lost if another actor were substituted?

Code	Label	Operational rule
AD-L	Low	the act’s substantive content is largely transferable or reproducible from public documents, generic arguments, or institutionally interchangeable evidence
AD-M	Moderate	some role-specific, local, professional, experiential, or organizational content would be lost, but substantial content remains substitutable
AD-H	High	the act depends materially on first-person experience, unique local/property exposure, particular standing, specialized role knowledge, or another actor-specific relation that substitution would erase

AD is not authority, influence, status, or importance. A resident may be AD-H with no formal power; a binding authority may make an AD-L or AD-M act.

B.8 Institutional configuration fields kept separate from the taxonomy and Grammar

Packet A also records, where observable:

• recipient;
• channel;
• formality;
• visibility;
• transmission;
• solicitation;
• record status;
• subject authority;
• route to binding authority;
• channel control;
• duty / required procedure;
• revisability;
• AI semantic fields;
• scope / realization referent.

These fields can explain conversion from act to consequence, but they do not replace RO/FP/AD.

B.9 Primary-signature rule

For descriptive occupancy only, one frozen primary signature is indexed for each Present confirmatory project. This does not convert multi-act projects into one-act cases and is not an act-frequency estimator.

Observed primary-signature occupancy among the 16 Present confirmatory projects:

Field	Code / label	Count
Participation Mode	M5 Contestatory / Claim-making	13
	M4 Deliberative / Consultative	1
	M1 Expressive / Discursive	1
	M8 Authoritative / Constitutive	1
RO	RO-C	10
	RO-O	4
	RO-R	1
	RO-S	1
FP	FP-O	10
	FP-F	4
	FP-C	2
AD	AD-L	10
	AD-M	3
	AD-H	3

B.10 Anti-shortcut rules

Do not infer:

• RO from Participation Mode;
• FP from generic project stage;
• AD from prominence or authority;
• participation from affectedness;
• implicit participation from mere silence, passive exposure, or unexpressed preference;
• tangible participation from consequence magnitude;
• intangible participation from presumed weakness;
• AI concern from project identity;
• capability-escalation fear from the mere word “AI”;
• consequence from the existence of an act;
• realization change from acknowledgment or procedure entry.

Appendix C — Frame One Denominator, Randomization, and Confirmatory Matrix

C.1 Denominator provenance

The canonical denominator is fixed at:

The complete original F1-to-project crosswalk did not survive in one intact artifact.

• 11 mappings were directly recovered.
• 38 mappings were mechanically reconstructed.
• Reconstruction rule: state alphabetical → frozen site/location alphabetical → sponsor/developer as same-location tie-break.
• The rule reproduces all directly recovered anchors.
• The original written ordering instruction was not recovered verbatim.
• The 38 reconstructed rows are never described as directly recovered originals.

This provenance issue affects identifier reconstruction, not the denominator size, random seeds, or sampled F1 positions.

C.2 Frozen 49-project crosswalk

F1 ID	Sponsor / Developer	Project	Frozen location	State	Mapping	Draw role	Notes
F1-001	Meta	Montgomery	Montgomery	AL	RECONSTRUCTED_V1	Confirmatory
F1-002	Google	West Memphis / Proctor	West Memphis	AR	DIRECT_RECOVERED	Pilot
F1-003	Microsoft	Fairwater Atlanta	Atlanta	GA	RECONSTRUCTED_V1	Neither
F1-004	AWS	Georgia infrastructure — Butts & Douglas Counties	Butts / Douglas Counties	GA	RECONSTRUCTED_V1	Neither	One sponsor-announcement project row spanning the two counties.
F1-005	Core Scientific	Dalton 4	Dalton	GA	RECONSTRUCTED_V1	Neither
F1-006	Google	Cedar Rapids	Cedar Rapids	IA	RECONSTRUCTED_V1	Neither
F1-007	Google	Fort Wayne	Fort Wayne	IN	RECONSTRUCTED_V1	Neither
F1-008	Meta	Jeffersonville	Jeffersonville	IN	RECONSTRUCTED_V1	Confirmatory
F1-009	Meta	Richland Parish / Hyperion	Richland Parish	LA	RECONSTRUCTED_V1	Confirmatory
F1-010	MillCompute	Bates Mill 3	Lewiston	ME	RECONSTRUCTED_V1	Neither
F1-011	OpenAI / Stargate	Saline Township / The Barn	Saline Township	MI	RECONSTRUCTED_V1	Neither
F1-012	Meta	Rosemount	Rosemount	MN	DIRECT_RECOVERED	Confirmatory	Resident-participation case previously coded under this slot.
F1-013	AWS	Madison County data-center complexes	Madison County	MS	RECONSTRUCTED_V1	Confirmatory	Two complexes treated as one project row in the frozen census.
F1-014	AWS	Warren County / Vicksburg	Warren County / Vicksburg	MS	DIRECT_RECOVERED	Pilot
F1-015	WhiteFiber	NC-1	Madison / Rockingham County	NC	DIRECT_RECOVERED	Pilot
F1-016	AWS	Richmond County	Richmond County	NC	RECONSTRUCTED_V1	Confirmatory
F1-017	Applied Digital	Polaris Forge 2	Harwood	ND	RECONSTRUCTED_V1	Confirmatory	Final-12 addition recovered by name.
F1-018	CoreWeave	Kenilworth	Kenilworth	NJ	DIRECT_RECOVERED	Pilot
F1-019	OpenAI / Stargate	Doña Ana County	Doña Ana County	NM	RECONSTRUCTED_V1	Neither
F1-020	AWS	Fayette County	Fayette County	OH	RECONSTRUCTED_V1	Confirmatory
F1-021	OpenAI / Stargate	Lordstown	Lordstown	OH	RECONSTRUCTED_V1	Confirmatory
F1-022	Core Scientific / CoreWeave	Muskogee	Muskogee	OK	DIRECT_RECOVERED	Pilot	Final-12 addition recovered by name.
F1-023	Google	Stillwater	Stillwater	OK	RECONSTRUCTED_V1	Neither
F1-024	AWS	Falls Township	Falls Township	PA	DIRECT_RECOVERED	Pilot
F1-025	CoreWeave	Lancaster	Lancaster	PA	RECONSTRUCTED_V1	Confirmatory
F1-026	AWS	Salem Township	Salem Township	PA	RECONSTRUCTED_V1	Neither
F1-027	Meta	Aiken County	Aiken County	SC	RECONSTRUCTED_V1	Confirmatory
F1-028	Google	Pine Hill Business Campus	Ridgeville	SC	RECONSTRUCTED_V1	Neither
F1-029	Google	Winding Woods Commerce Park	St. George	SC	RECONSTRUCTED_V1	Neither
F1-030	xAI	Colossus II / Tulane Road	Memphis	TN	DIRECT_RECOVERED	Pilot
F1-031	TeraWulf / Fluidstack	Abernathy AI data-center project	Abernathy	TX	RECONSTRUCTED_V1	Confirmatory	Final-12 addition recovered by name.
F1-032	OpenAI / Stargate / Crusoe	Stargate Abilene	Abilene	TX	RECONSTRUCTED_V1	Confirmatory
F1-033	Google	Armstrong County campus	Armstrong County	TX	RECONSTRUCTED_V1	Confirmatory
F1-034	Cipher / Fluidstack	Barber Lake	Barber Lake / Colorado City	TX	RECONSTRUCTED_V1	Neither	Final-12 addition recovered by name.
F1-035	Nscale / Ionic	Cedarvale	Cedarvale / Barstow / Ward County	TX	RECONSTRUCTED_V1	Confirmatory	Final-12 addition recovered by name.
F1-036	Priority Power	Project Aggie	College Station	TX	RECONSTRUCTED_V1	Neither	Final-12 addition recovered by name.
F1-037	Galaxy / CoreWeave	Helios	Dickens County	TX	DIRECT_RECOVERED	Pilot	Final-12 addition recovered by name.
F1-038	Meta	El Paso AI-optimized data center	El Paso	TX	RECONSTRUCTED_V1	Confirmatory
F1-039	Google	Haskell County campus	Haskell County	TX	RECONSTRUCTED_V1	Confirmatory
F1-040	OpenAI / Stargate / SB Energy	Stargate Milam County	Milam County	TX	RECONSTRUCTED_V1	Neither
F1-041	CoreWeave / Poolside	Project Horizon / Longfellow Ranch	Pecos County	TX	DIRECT_RECOVERED	Pilot
F1-042	Fermi America	HyperGrid / Project Matador	Project Matador / Texas Panhandle	TX	RECONSTRUCTED_V1	Confirmatory	Final-12 addition recovered by name.
F1-043	OpenAI / Stargate	Stargate — Shackelford County	Shackelford County	TX	DIRECT_RECOVERED	Pilot	Later identity audit confirms the same underlying physical campus as F1-044 Frontier; retained as the frozen pilot sponsor-announcement row.
F1-044	Vantage Data Centers	Frontier	Shackelford County	TX	RECONSTRUCTED_V1	Confirmatory	Later identity audit confirms the same underlying physical campus as F1-043; retained under frozen confirmatory randomization with leave-one-out sensitivity.
F1-045	Google	Chesterfield County	Chesterfield County	VA	RECONSTRUCTED_V1	Neither
F1-046	Meta	Beaver Dam	Beaver Dam	WI	RECONSTRUCTED_V1	Neither
F1-047	Microsoft	Fairwater / Mount Pleasant	Mount Pleasant	WI	RECONSTRUCTED_V1	Confirmatory
F1-048	OpenAI / Stargate / Vantage	Port Washington	Port Washington	WI	RECONSTRUCTED_V1	Confirmatory
F1-049	Meta	Cheyenne	Cheyenne	WY	RECONSTRUCTED_V1	Neither

C.3 Frozen draws

Pilot: seed 20260828,

F1-002, F1-014, F1-015, F1-018, F1-022, F1-024, F1-030, F1-037, F1-041, F1-043

Pilot cases are instrument-development only and excluded from confirmatory estimates.

Confirmatory: seed 20260829,

F1-001, F1-008, F1-009, F1-012, F1-013, F1-016, F1-017, F1-020, F1-021, F1-025, F1-027, F1-031, F1-032, F1-033, F1-035, F1-038, F1-039, F1-042, F1-044, F1-047, F1-048

No redraw, substitution, or reordering is permitted because a case is difficult, silent, retrieval-limited, right-censored, or unusually salient.

C.4 Confirmatory status at the August 28, 2026 research cutoff

F1 ID	Project	Packet A state at cutoff	Packet B maturity	Frozen endpoint if right-censored
F1-001	Meta Montgomery Data Center — Montgomery, Alabama	Unknown	Mature	—
F1-008	Meta Jeffersonville Data Center — Jeffersonville, Indiana	Present	Mature	—
F1-009	Meta Richland Parish / Hyperion Data Center — Richland Parish, Louisiana	Present	Mature	—
F1-012	Meta Rosemount Data Center — Rosemount, Minnesota	SearchedNotFound	Mature	—
F1-013	AWS Madison County Data-Center Complexes — Madison County, Mississippi	Present	Mature	—
F1-016	AWS Energy Way Tech Campus — Richmond County, North Carolina	Present	Mature	—
F1-017	Applied Digital Polaris Forge 2 — Harwood, North Dakota	Present	Mature	—
F1-020	AWS Fayette County / Jefferson Township Data Center — Ohio	SearchedNotFound	Mature	—
F1-021	OpenAI / SoftBank Stargate — Lordstown, Ohio	Present	Right-censored	2026-09-23
F1-025	CoreWeave Lancaster AI Data Center — Lancaster, Pennsylvania	Present	Mature	—
F1-027	Meta Aiken County Data Center — South Carolina	Present	Mature	—
F1-031	TeraWulf / Fluidstack Abernathy AI Data Center — Hale County, Texas	Unknown	Right-censored	2026-10-28
F1-032	Stargate Abilene AI Data Center — Abilene / Taylor County, Texas	Present	Mature	—
F1-033	Google / Goodnight AI Data-Center Campus — Armstrong County, Texas	Present	Right-censored	2026-11-14
F1-035	Nscale / Ionic Cedarvale AI Data Center — Ward County, Texas	Unknown	Right-censored	2026-10-14
F1-038	Meta El Paso AI Data Center — El Paso, Texas	Present	Right-censored	2026-10-15
F1-039	Google Haskell County AI Data Centers — Texas	Present	Right-censored	2026-11-14
F1-042	Fermi America HyperGrid / Project Matador — Texas Panhandle	Present	Mature	—
F1-044	Vantage Data Centers — Frontier, Shackelford County, Texas	Present	Mature	—
F1-047	Microsoft Fairwater — Mount Pleasant, Wisconsin	Present	Right-censored	2026-09-18
F1-048	OpenAI / Stargate / Vantage — Port Washington, Wisconsin	Present	Mature	—

C.5 Unit firewall

The frozen unit is the sponsor/project row used in the denominator. Later evidence can reveal overlap between announcement identity and physical-campus identity.

F1-043 and F1-044 are the controlling example: – F1-043 — OpenAI/Stargate, Shackelford County — pilot; – F1-044 — Vantage Frontier, Shackelford County — confirmatory.

Later evidence connects the announcements to the same underlying physical campus. The rows remain frozen because randomization already occurred, but there is no confirmatory duplicate contribution because F1-043 is pilot-excluded.

C.6 Scope classes preserved inside cases

Case records distinguish: – focal project; – linked infrastructure; – future expansion; – same-place/regional project; – second-order governance.

Same-place or linked-infrastructure participation is not automatically focal-project participation.

Appendix D — Frozen Packet B Consequence and Realization Protocol

The text below is the frozen Packet B protocol used for the empirical work. It is reproduced here so the outcome/consequence boundary is inspectable without relying on the manuscript narrative.

Frame One — Packet B Coding Protocol v1.0

Status: FROZEN before coding the first Packet-B case with an observed qualifying participatory act. Inheritance: Direct operationalization of Gradual AGI #9, §6 consequence classes and the already-frozen Frame One t0+12-month design. Rule: Do not revise silently after outcome inspection.

1. Separation rule

Packet B contains two distinct objects:

1. Project realization status at the frozen endpoint — what happened to the focal realization by t0 + 12 months.
2. Participatory Consequence — what change, persistence, or absence of detected change is temporally and substantively associated with an observed Participatory Act / Configuration.

The first can be coded even when no qualifying act is observed. The second cannot be inferred from project outcome alone.

2. Frozen observation window

Use the same project-specific interval frozen in Packet A: [t0, t0 + 12 months].

Later evidence may only corroborate what had already occurred by the endpoint; later changes are post-window metadata.

For right-censored cases: – initialize Packet B if useful; – mark relevant fields RightCensored; – do not finalize the 12-month outcome before the endpoint.

3. Project realization status

Allowed primary vocabulary: – Proceeding – Conditioned – Delayed – Relocated – Reduced – Denied – Withdrawn – NotYetOperationalButProceeding – Unresolved – RightCensored

Multiple descriptors may be retained when necessary. Do not impose an ordinal scale.

4. Consequence assessability

Packet A = Present

Participatory Consequence is in principle assessable, subject to source visibility.

Packet A = SearchedNotFound

Use ParticipatoryConsequenceAssessability = NotApplicable_NoDetectedAct.

Do not call this a Null consequence. The published Null class presupposes a documented act/configuration followed by no detected consequence.

Packet A = Unknown

Use ParticipatoryConsequenceAssessability = NotAssessable_ParticipationUnknown.

Project realization may still be coded independently.

5. Consequence classes

Preserve the seven published #9 classes:

1. Informational
2. Procedural
3. InstitutionalReturnPath
4. SubjectReadiness
5. Distributional
6. RealizationLevel
7. NullNoDetectedChange

For each class use: – Present – NoDetectedChange – Unresolved – NotAssessable – NotApplicable

Classes are non-exclusive and not ordered.

NullNoDetectedChange = Present only when: – a qualifying act/configuration is documented; – the frozen consequence search is completed through the endpoint; – no informational, procedural, institutional, readiness, distributional, or realization-level change linked to that act/configuration is detected.

6. Uptake ladder — descriptive, not scalar

Record separately: – RecipientReached – ProcedureEntry – AcknowledgmentOrResponse – ReasonedResponse – DecisionChange – RealizationChange

Use Yes / No / Unresolved / NotApplicable. Never sum these into a score.

7. Linkage-support grade

Grades support for association, not causal effect: – L0 None/Unassessable – L1 TemporalSequenceOnly – L2 ProceduralOrSubstantiveLinkage – L3 DirectDocumentaryAttribution – L4 DirectAttributionPlusIndependentCorroboration

Higher grade does not establish sole causation.

8. Competing explanations

Every non-null consequence claim must record visible non-participatory drivers such as law/regulation, commercial incentives, infrastructure constraints, technical feasibility, formal authority, litigation risk, developer strategy, market demand, and institutional routine.

9. Responsive Floor Test boundary

Do not label a case a Responsive Floor merely because participation and a consequence coexist. Responsive Floor is a later cross-case test of whether a pre-specified participatory feature contributes to realization differences under sufficiently comparable other conditions.

10. Version rule

If new evidence changes a Packet-B field: – preserve the earlier packet; – issue a new version; – state what evidence changed the field.

Packet-B outcomes may never be used to rewrite Packet-A Mode/RO/FP/AD retrospectively.

Appendix E — Case Evidence Supplement

The full 21-case Packet A/Packet B evidence and primary-source register is published separately as Supplement S1 — Confirmatory Case Evidence and Primary-Source Register. This keeps the publication-facing appendices readable without removing the evidentiary backplane.

Appendix F — Reliability and Reproducibility Provenance

Status: reliability statistic remains NotReconstructable_ArchivalLoss.

1. What survives

One complete assignment-blinded X01–X08 coding return survives. It records, for every holdout: – Inclusion; – primary Participation Mode / internal M-code; – Realization Orientation; – Foreclosure Position; – coder ambiguity notes.

It is a valid one-coder raw record and demonstrates that the instrument was actually used outcome-blind, including use of unresolved FP rather than forced resolution.

2. What the matching paired archive does not safely preserve

Required provenance element	Status
exact coder identity / coder class for the other trace	not safely established
exact frozen packet version	not safely established
exact target / rule version	not safely established
complete row-by-row raw categorical fields	incomplete
X02 Foreclosure Position	missing at source resolution
proof all fields precede any rule repair/adjudication	not safely established
Actor Dependence assignments, if AD belonged to that holdout version	not preserved / uncertain

These are not cosmetic metadata. Version/timing uncertainty can make two coding tables non-comparable under the registered protocol.

3. Why this happened

The preserved archive establishes the condition of the loss, not a specific historical cause for it.

The project should therefore not claim: – a coder left the project; – a software failure occurred; – a particular file was accidentally deleted; – or another unrecorded causal story.

The defensible statement is:

The research workflow preserved one complete blind return and an incomplete trace of another coding return, but did not preserve the complete matching pre-adjudication pair with sufficient coder/version/target provenance for registered reliability reconstruction.

This is a data-retention / provenance failure at the paired-archive level, with the precise historical failure event undocumented.

4. Why no coefficient can be reconstructed

A coefficient could be manufactured only by assuming: – that the incomplete trace is the matching first substantive coder; – both coders used the same packet and rule versions; – a value for missing X02 FP; – the timing of rule repair/adjudication; – and, where relevant, AD values.

Those assumptions violate the frozen protocol.

Therefore:

PropositionA_Reliability = NotReconstructable_ArchivalLoss

not pass, fail, zero reliability, or alpha below threshold.

5. What the surviving holdout does show

It supports only these bounded statements:

1. assignment-blinded coding was executed;
2. at least one coder could apply the instrument without outcome information;
3. the coder used unresolved FP where the packet did not support a forced state;
4. ambiguity occurred at identifiable boundaries;
5. the holdout was deliberately heterogeneous across Participation Modes.

It does not establish reproducibility, field-level agreement, an alpha threshold, or validated RO/FP/AD measurement.

6. Epistemic calibration

Preferred reader-facing result:

Relational Grammar: discrimination among RO/FP/AD states is observed in the executed outcome-blind coding; the surviving archive supports no inter-coder reproducibility or instrument-validation claim. The observation is hypothesis-generating pending independent replication.

Internal codes remain: – PropositionA_Nonredundancy = DiscriminantEvidenceObserved – PropositionA_Reliability = NotReconstructable_ArchivalLoss – PropositionA_Codability = ExercisedButReliabilityUnresolved

The phrase “partial support” should appear only if immediately defined as nonredundancy-only and provisional pending properly archived reliability replication.

7. Replication requirement

A valid replication should preserve immutable packet version/hash, coder identity/role, target/codebook version, complete pre-adjudication assignments, missing/unresolved values, adjudication as a separate layer, and analysis code for raw agreement and Krippendorff alpha.

No further reconstruction should be attempted by guessing missing values or pairing different workflow traces.

Appendix G — Temporal, Sentiment, Media, and Institutionalization Specifications

G.1 Firewall among empirical layers

The temporal/context extension keeps four objects separate:

None is pooled into a common negative-sentiment or resistance index.

G.2 Focal resistance-event frequency

Primary descriptive exposure: – pre-2026 exposure: 4,632 project-days = 152.1807 month-equivalents; – 2026 through August 28: 2,697 project-days = 88.6078 month-equivalents.

Broad focal resistance episodes: – pre-2026: 23; – 2026 through August 28: 10.

Rates: – pre-2026: 15.1136 events per 100 project-months; – 2026: 11.2857 per 100 project-months.

Crude rate ratio:

Approximate Poisson 95% CI:

with .

Canonical verdict:

Calendar2026ResistanceFrequencyIncrease = NotSupported.

This is not a finding that resistance declined.

G.3 Age-adjusted sensitivity models

Poisson event-count model with exposure offset — all cases

• calendar 2026 IRR = 0.6754; 95% CI 0.2506–1.8205;
• project age +12 months IRR = 1.6666; 95% CI 0.2779–9.9935;

Poisson model excluding F1-001

• calendar 2026 IRR = 0.6157; 95% CI 0.2303–1.6461;
• project age +12 months IRR = 1.7377; 95% CI 0.2932–10.3008;

Binary any-event project-month model

• calendar 2026 OR = 0.6318; 95% CI 0.2005–1.9904;
• project age +12 months OR = 2.6254; 95% CI 0.4356–15.8221;

No positive calendar-year or project-age frequency effect is statistically identified.

G.4 Formal-entry model

Across 33 focal/focal-oversight broad resistance episodes:

• calendar 2026 OR = 0.5328; 95% CI 0.1093–2.5978;
• project age +12 months OR = 0.3583; 95% CI 0.0200–6.4096;

Descriptive formal/institutional entry is approximately: – 69.6% pre-2026; – 50.0% in 2026.

Thus there is no identified late-stage increase in formal entry itself.

G.5 Downstream institutionalization sensitivity

The candidate maturation mechanism concerns downstream conversion into conditions, mitigation, CBAs, moratoria, reporting duties, return paths, or related institutional adaptation.

Exploratory sensitivities:

• mixed-scope first-resistance age +1 month OR = 1.4509; 95% CI 0.8940–2.3550;
• strict focal OR = 1.2724; 95% CI 0.8280–1.9554;
• strict focal + precise date OR = 1.2202; 95% CI 0.7898–1.8850;

Point estimates are positive but imprecise and weaken under stricter coding; the temporal data are insufficient for a maturation inference.

Canonical verdict:

InstitutionalizationMaturation = NotIdentified_TemporalDataInsufficient.

The manuscript’s temporal interpretation is therefore:

G.6 Survey context series

Survey evidence is context only.

Pew — 2025

Survey fielded June 9–15, 2025: – 50% more concerned than excited about increasing AI use; – 10% more excited than concerned; – 57% rated societal risks as high; – roughly six in ten wanted more control.

Interpretation: wary AI sentiment predates 2026.

Quinnipiac — March 2026

Fielded March 19–23: – 65% opposed an AI data center in their community; – 24% supported; – among opponents: 72% electricity cost, 64% water use, 41% noise.

Interpretation: strong local opposition can coexist with infrastructure-centered concern.

Pew — 2026

Survey fielded February 17–23, published June 17: – 67% had little/no confidence in government to regulate AI effectively, versus 62% in 2024; – 59% lacked confidence in U.S. companies to develop/use AI responsibly.

Reuters/Ipsos — June 2026

• 57% opposed a data center in their own community;
• 33% approved of the rapid construction pace supporting AI;
• electricity, land, water, limited long-term jobs, transparency, and economic distribution were prominent.

No pooled sentiment index is constructed.

G.7 Media and social-discourse protocols

The MC-N mainstream-news and MC-S social-platform protocols were designed to separate: – attention; – evaluation; – concern basis; – amplification/mobilization.

The quantitative retrieval/coding corpus was not executed to completion for this manuscript.

Therefore: – no quantitative MC-N result is reported; – no cross-platform MC-S prevalence result is reported; – external monitoring reports can serve only as hypothesis/context benchmarks.

The Soufan Center’s reported 2026 rise in anti-data-center narratives is not substituted for Frame One because its complete query, platform access, and sampling denominator were not independently reproduced.

G.8 AI-semantic mechanism fields

Keep separate: – AIContentPresent; – AISpecificConcernPresent; – AICapabilityEscalationFear; – ConcernBasis.

No focal coded temporal event establishes:

AICapabilityEscalationFear = Yes.

Canonical verdict:

AICapabilityEscalationFearMechanism = NotSupportedInFrameOne.

This does not mean broader public AI anxiety failed to rise. It means the focal data-center evidence does not identify stronger-capability fear as the mechanism behind the observed resistance pattern.

G.9 Context use rule

Contextual surveys, media material, elite-response examples, and external monitoring may: – characterize the surrounding attitude environment; – suggest alternative mechanisms; – identify hypotheses for later studies.

They may not: – determine Packet A participation coding; – change Frame One membership; – convert infrastructure concern into AI fear; – establish project-level causal attribution; – or rescue an unsupported calendar-frequency claim.

Appendix H — Pilot Cases and Archive Status

H.1 Frozen pilot draw

Seed: 20260828
Pilot N = 10

F1	Sponsor / developer	Project	Location	State	Mapping status
F1-002	Google	West Memphis / Proctor	West Memphis	AR	DIRECT_RECOVERED
F1-014	AWS	Warren County / Vicksburg	Warren County / Vicksburg	MS	DIRECT_RECOVERED
F1-015	WhiteFiber	NC-1	Madison / Rockingham County	NC	DIRECT_RECOVERED
F1-018	CoreWeave	Kenilworth	Kenilworth	NJ	DIRECT_RECOVERED
F1-022	Core Scientific / CoreWeave	Muskogee	Muskogee	OK	DIRECT_RECOVERED
F1-024	AWS	Falls Township	Falls Township	PA	DIRECT_RECOVERED
F1-030	xAI	Colossus II / Tulane Road	Memphis	TN	DIRECT_RECOVERED
F1-037	Galaxy / CoreWeave	Helios	Dickens County	TX	DIRECT_RECOVERED
F1-041	CoreWeave / Poolside	Project Horizon / Longfellow Ranch	Pecos County	TX	DIRECT_RECOVERED
F1-043	OpenAI / Stargate	Stargate — Shackelford County	Shackelford County	TX	DIRECT_RECOVERED

All ten frozen pilot mappings are DIRECT_RECOVERED in the canonical denominator ledger.

H.2 Role of the pilot

The pilot was used for instrument development before the separate confirmatory draw. Pilot rows are excluded from all primary confirmatory estimates and are not pooled with the confirmatory sample.

The pilot served to expose coding-boundary, scope/focality, missing-data/search-state, and anti-rescue issues before confirmatory execution.

H.3 Pilot/confirmatory consistency

A project-level pilot Packet A/Packet B result archive sufficient to recompute final-codebook: – ParticipationEvidence distributions; – Participation Mode / RO / FP / AD occupancy; – consequence-class counts; – realization outcomes;

was not located in the preserved project archive used for this mechanical check.

Accordingly:

PilotConfirmatoryConsistency = NotAssessable_ArchiveInsufficient

This is not evidence of inconsistency. It means the pilot cannot now be used as a retrospective validation set without newly recoding the ten development cases under the final instrument.

That recoding is not performed, because it would occur after confirmatory results are known and would blur the development/confirmation boundary.

H.4 Unit note

F1-043 is the pilot OpenAI/Stargate Shackelford County announcement row. Later evidence connects it to the same underlying physical campus as confirmatory F1-044 Vantage Frontier.

The frozen IDs remain unchanged. F1-043 is pilot-excluded, so it does not duplicate the physical campus inside the confirmatory estimate.

H.5 Manuscript use

This appendix answers which cases constituted the pilot and what role they played. It should not claim a pilot/confirmatory consistency result that the surviving archive cannot support.

H.6 Pilot-to-confirmatory instrument status

The pre-pilot protocol allowed a load-bearing field or retrieval procedure to be revised before confirmatory coding if a frozen pilot gate required repair; such changes had to be logged and versioned. The surviving archive does not support a complete final-codebook pilot reanalysis or an exhaustive itemized repair ledger.

Accordingly, the pilot is described as instrument development, the confirmatory instrument is the operative instrument for confirmatory cases, and the pilot is not claimed to validate the final instrument. No retrospective pilot recoding is performed after confirmatory outcomes are known.

Appendix I — Frame-Provenance Sensitivities

This appendix preserves the principal frame-provenance and leave-one-out checks. None changes the frozen denominator or confirmatory draw.

I.1 F1-044 same-campus / reconstructed-row sensitivity

Purpose: address the later discovery that pilot F1-043 and confirmatory F1-044 are frozen sponsor-announcement rows referring to the same underlying physical campus.

Rule: sensitivity only. F1-044 remains in the frozen confirmatory draw and is not renumbered, substituted, or deleted.

1. Why F1-044 is the sensitivity target

• F1-043 is pilot-only and directly recovered.
• F1-044 is confirmatory and mechanically reconstructed.
• Later evidence identifies both announcement rows with the same underlying Shackelford County campus.
• Because F1-043 is excluded from confirmatory estimates, the primary confirmatory analysis does not count the physical campus twice.
• The sensitivity temporarily excludes F1-044 from calculations while leaving the frozen sample untouched.

2. Case-flow sensitivity

Primary cutoff: 21 = 14 mature + 7 right-censored.

F1-044-excluded sensitivity: 20 = 13 mature + 7 right-censored, with 13 = 10 Present + 2 SearchedNotFound + 1 Unknown.

3. Grammar occupancy

F1-044 contributes the only primary-signature M8 / RO-S case.

Field	Full frame	Excluding F1-044
Participation Mode	M5=13, M4=1, M1=1, M8=1	M5=13, M4=1, M1=1
RO	C=10, O=4, R=1, S=1	C=10, O=4, R=1
FP	O=10, F=4, C=2	O=9, F=4, C=2
AD	L=10, M=3, H=3	L=10, M=2, H=3

Implication: the statement that all four RO values appear in primary-signature occupancy is sensitive to F1-044. The stronger nonredundancy evidence is not: same-form/different-RO contrasts, within-project FP variation, and AD variation remain without F1-044.

4. Mature consequence counts

Consequence classes are non-exclusive.

Consequence	Full 11 Present	Excluding F1-044: 10 Present
Informational Present	8	8
Procedural Present	5	4
Institutional/return-path Present	3	3
Distributional Present	1	1
Realization-level Present	3	2
Null Present	1	1

Strongest linkage: – full: L3=6, L2=3, L0=2; – excluding F1-044: L3=5, L2=3, L0=2.

Implication: informational uptake remains much more common than realization-level change (8/10 vs 2/10). The two remaining realization-level cases are F1-025 and F1-048, both non-veto configurations. The direct-authority symmetry/comparator case is what disappears.

5. Institutional contrasts

Mechanical exclusion yields: – Contrast A required-procedure cell: n=8; ProcedureEntry 8/8; uptake 3 Yes / 4 No / 1 Unresolved. The absent/not-established cell remains n=2, 0/2 entry, 0/2 uptake. – Contrast C clear binding-authority route: n=8; entry 8/8; uptake 3 Yes / 4 No / 1 Unresolved. Sparse comparator cells are unchanged. – Contrast D open primary referent: n=8; uptake 2 Yes / 5 No / 1 Unresolved. All three remaining uptake-positive configurations (F1-009, F1-025, F1-048) still contain an open/later-open pathway. – Contrast E: Present n=10 gives 5 no material alteration and 5 conditioned/altered; SearchedNotFound remains n=2 with 2 no alteration and 0 conditioned/altered.

The frozen precision gate remains unmet. No aggregate inference becomes admissible.

6. Non-veto / Responsive Floor result

Full mature realization-level set: – F1-025 — non-veto, L2; – F1-044 — direct authority, L3; – F1-048 — non-veto, L3.

F1-044-excluded set: – F1-025 — non-veto, L2; – F1-048 — non-veto, L3.

Thus the core existence result is not dependent on F1-044. What is lost is the direct-authority contrast.

The corresponding sensitivity statement is: > The documented mediated non-veto association sequence remains present when the reconstructed F1-044 row is excluded in sensitivity analysis; the direct-authority comparator does not.

ResponsiveFloor_FrameOne = NotAdjudicated remains unchanged.

7. Temporal-frequency sensitivity

F1-044 contributes no focal broad-resistance event to the frozen event count. It contributes exposure only: – pre-2026: 135 days; – 2026 through Aug. 28: 231 days.

Excluding F1-044: – pre-2026 exposure = 4497 days = 147.7454 month-equivalents; – 2026 exposure = 2466 days = 81.0185 month-equivalents; – event counts remain 23 and 10; – rates become 15.5673 vs 12.3429 per 100 project-months; – crude RR = 0.7929; – approximate 95% CI = 0.377–1.666; p≈0.540.

Age-adjusted Poisson, same frozen specification: – calendar 2026 IRR = 0.7265, 95% CI 0.2749–1.9197, p=0.519; – project age +12 months IRR = 1.6126, 95% CI 0.2689–9.6706, p=0.601.

Binary any-event-month sensitivity: – calendar 2026 OR = 0.6884, 95% CI 0.2232–2.1236, p=0.516; – project age +12 months OR = 2.5375, 95% CI 0.4211–15.2914, p=0.310.

F1-044 contributes no broad-resistance event to the formalization-event model and no row to the downstream institutionalization-latency test. Those analyses are unchanged.

Calendar2026ResistanceFrequencyIncrease = NotSupported remains unchanged.

8. Sensitivity verdict

F1_044_ExclusionSensitivity = HeadlineConclusionsStable

Changed: – sole primary-signature RO-S / M8 occupancy disappears; – Procedural Present 5→4; – RealizationLevel Present 3→2; – L3 count 6→5; – direct-authority comparator removed; – exposure-adjusted temporal estimates shift modestly toward 1.

Unchanged: – Grammar nonredundancy evidence; – Act ⇏ Consequence ⇏ Realization separation; – documented L3 mediated non-veto association sequence (F1-048); – sparse/insufficient Contrast E; – Responsive Floor non-adjudication; – no 2026 resistance-frequency increase; – capability-fear verdict; – no-scalar formal verdict.

I.2 Confirmatory AI-gate composition

Question: did the protocol’s governance-side AI-relevance alternative materially select the 21-case confirmatory sample?

Result

Mechanical review of the frozen 21 Packet A files finds:

• 20 / 21 explicitly state that the project-side AI gate passes / is satisfied.
• 1 / 21 — F1-013 AWS Madison County is coded PASS — contextual/public AI linkage with Moderate confidence; the packet explicitly notes that the sponsor announcement itself emphasized hyperscale/cloud infrastructure rather than a Meta-style AI workload claim.
• 0 / 21 is included on a governance-side-only AI gate in the preserved Packet A records.
• The Packet A files do not systematically code a separate “both project-side and governance-side” flag once project/public AI relevance is satisfied, so a Both count would require retrospective reinterpretation and is not reported.

Implication

The protocol permits a governance-side AI-relevance path, but that path is not the sole inclusion basis for any confirmatory case in the executed draw. The specific concern that confirmatory membership was populated because governmental actors had already labeled projects AI-related is therefore not supported by the preserved Packet A gate records.

A narrower limitation remains: – AI relevance is heterogeneous in strength; – F1-013 has weaker/contextual provenance; – publicly AI-labeled infrastructure may differ from less publicly labeled infrastructure.

The corresponding AI-gate statement is:

The inclusion protocol allowed either project/public AI relevance or governance-side AI-specific relevance. In the executed confirmatory draw, however, no case relies on governance-side relevance alone: 20 of 21 Packet A records explicitly pass a project-side AI gate, and the remaining case, F1-013, passes on a weaker contextual/public AI linkage. Governance-side-only qualification therefore does not drive confirmatory membership, although AI-relevance provenance is heterogeneous across projects.

GovernanceSideOnlyConfirmatoryCases = 0

AIGateComposition = 20_ExplicitProjectSide + 1_ContextualPublic + 0_GovernanceSideOnly

I.3 Full denominator campus-identity audit

A complete mechanical screen evaluated all =1,176 denominator pairs using state, normalized location, project-name similarity, and sponsor/project overlap. Four pairs were flagged: F1-043/F1-044 by exact Shackelford County location, and three pairs among Abilene, Milam County, and Shackelford County because of shared Stargate branding.

External verification confirms F1-043/F1-044 as one physical Frontier/Stargate campus. Abilene, Milam County, and Shackelford County are separately identified Stargate sites. A secondary review of alias-prone tenant/developer, landlord/tenant, phased-campus, and repeated-brand rows found no additional cross-row physical-campus collision.

DenominatorCampusIdentityAudit = OneConfirmedCollision_F1-043_F1-044

AdditionalCampusCollisionsIdentified = 0

F1-044 remains primary because the sampled unit is the frozen sponsor/project announcement row; deletion after randomization would alter the confirmatory sample. F1-043 is pilot-only, so the physical campus is represented once in confirmatory estimates. Appendix I.1 supplies the leave-one-out sensitivity.

I.4 F1-013 AI-relevance boundary and leave-one-out sensitivity

F1-013 is the only confirmatory case whose sponsor announcement did not itself name an AI workload. Its Moderate-confidence gate rests on contextual/public evidence available independently of participatory outcomes: same-day Reuters coverage linked the expansion to AI-driven cloud demand, and later in-window public economic-development coverage explicitly linked the operation to AI and machine learning.

F1_013_AIGate = ContextualPublic_Moderate_NotOutcomeConditioned

Because the frozen Packet A itself required a sensitivity for this boundary, F1-013 is temporarily excluded without changing the frozen sample.

I.4.1 Primary-signature sensitivity

F1-013 contributes M5 · RO-C · FP-O · AD-L.

Field	Full 16 primary signatures	Excluding F1-013
Participation Mode	M5=13; M4=1; M1=1; M8=1	M5=12; M4=1; M1=1; M8=1
RO	C=10; O=4; R=1; S=1	C=9; O=4; R=1; S=1
FP	O=10; F=4; C=2	O=9; F=4; C=2
AD	L=10; M=3; H=3	L=9; M=3; H=3

No Mode or Grammar category disappears, and the same-Mode/different-Grammar contrasts remain.

I.4.2 Consequence and contrast sensitivity

Among mature Present cases, excluding F1-013 changes Informational Present 8→7 and L3 6→5. Procedural Present remains 5; ReturnPath Present remains 3; Distributional Present remains 1; RealizationLevel Present remains 3; Null remains 1. The three realization-level Present cases remain F1-025, F1-044, and F1-048.

For the registered descriptive contrasts, F1-013 removes one non-uptake observation from the required-procedure, clear-authority-route, and open-revisability cells. All A–D adjudication statuses remain unchanged. Contrast B remains unexercised.

Contrast E changes from Present n=11 (6 unchanged-path / 5 conditioned-or-altered) to Present n=10 (5 / 5), while SearchedNotFound remains n=2 (2 / 0). The pre-specified n≥20 per-cell precision gate remains far from met.

I.4.3 Responsive Floor and temporal sensitivity

The Port Washington L3 non-veto sequence is unaffected, and the aggregate Responsive Floor Test remains NotAdjudicated.

F1-013 contributes 367 pre-2026 exposure days and no focal broad-resistance event. Under the unchanged project-month specification, exclusion shifts the crude 2026/pre-2026 rate ratio from approximately 0.747 to 0.688. The age-adjusted calendar-2026 Poisson IRR shifts from approximately 0.675 to 0.607 (95% CI approximately 0.227–1.623; ). The binary any-event-month calendar-2026 OR shifts from approximately 0.632 to 0.557 (95% CI approximately 0.175–1.767; ). F1-013 contributes no row to the formalization-event or institutionalization-latency models, so those analyses are unchanged.

I.4.4 Verdict

F1_013_ExclusionSensitivity = HeadlineConclusionsStable

The sensitivity does not change Grammar discrimination observations, type→consequence contingency, the Port Washington documented non-veto association sequence, Contrast-E non-operability, aggregate Responsive Floor non-adjudication, the 2026 resistance-frequency verdict, the capability-fear result, or the no-scalar formal verdict.

Appendix J — Secondary Descriptive 3M Application Matrix

Status: post-freeze descriptive application to the 16 frozen ParticipationEvidence=Present primary acts. Modality/Participation Mode, RO, FP, and AD are carried forward unchanged from Packet A. Manifestness and Materiality are secondary descriptive assignments from the preserved act record; consequence was not used to choose them.

F1	Frozen primary act (short label)	Manifestness*	Participation Mode†	Materiality*	RO†	FP†	AD†
F1-008	IURC intervention/testimony	Explicit	M5	Mixed	RO-C	FP-O	AD-M
F1-009	LPSC motion	Explicit	M5	Mixed	RO-C	FP-O	AD-L
F1-013	legislative amendment	Explicit	M5	Mixed	RO-C	FP-O	AD-L
F1-016	stop-project petition	Explicit	M5	Intangible	RO-O	FP-F	AD-L
F1-017	project-specific hearing intervention	Explicit	M5	Intangible	RO-C	FP-O	AD-L
F1-021	email to planning official / council record	Explicit	M5	Intangible	RO-C	FP-F	AD-L
F1-025	City Council stop-project comment	Explicit	M5	Intangible	RO-O	FP-F	AD-L
F1-027	County Council public claim	Explicit	M5	Intangible	RO-C	FP-O	AD-L
F1-032	City Council opposition comment	Explicit	M5	Intangible	RO-O	FP-O	AD-M
F1-033	PUCT direct testimony	Explicit	M4	Intangible	RO-C	FP-O	AD-H
F1-038	pause-construction petition	Explicit	M5	Intangible	RO-C	FP-C	AD-L
F1-039	attributed lived-experience report	Explicit	M1	Intangible	RO-R	FP-F	AD-H
F1-042	anti-project protest/rally participation	Explicit	M5	Mixed	RO-O	FP-O	AD-L
F1-044	project-specific water-line authorization	Explicit	M8	Tangible	RO-S	FP-O	AD-M
F1-047	federal nuisance complaint	Explicit	M5	Mixed	RO-C	FP-C	AD-H
F1-048	annexation-hearing intervention	Explicit	M5	Intangible	RO-C	FP-O	AD-L

* Secondary descriptive code introduced after the original Packet-A freeze.
† Frozen/pre-existing Packet-A field reproduced without alteration.

J.1 Descriptive occupancy

• Manifestness: Explicit 16/16; Implicit 0/16.
• Materiality: Intangible 10/16; Mixed 5/16; Tangible 1/16.
• Participation Mode: M5 13/16; M4 1/16; M1 1/16; M8 1/16.

These counts are not prevalence estimates and are not used for effect-rate comparisons. The all-Explicit result is interpreted as evidence that the public-record observation strategy structurally favors overt participation.

J.2 Materiality precedence rule

ProcedureEntry=Yes does not automatically imply Mixed Materiality. Public comment or testimony can remain Intangible when formal procedure is merely the channel. Mixed is reserved for focal acts whose operative function itself includes a distinct legal/procedural or embodied component, such as a formal motion, legislative amendment, lawsuit, or protest. Materiality classifies the act, not its later consequence.

Appendix K — Contrast E Descriptive Inventory

Contrast E compares project realization outcomes for mature Present and SearchedNotFound cases without conditioning on consequence. It is retained for transparency but is not inferentially exercised because the frozen precision gate requires at least 20 determinate projects in each binary cell.

ParticipationEvidence	n	Proceeded with no material path alteration detected	Proceeded with material conditions/design/configuration change	Delayed	Relocated	Reduced	Denied	Withdrawn
Present	11	6	5	0	0	0	0	0
SearchedNotFound	2	2	0	0	0	0	0	0

The five Present projects in the conditioned/altered project-outcome category are F1-009, F1-025, F1-032, F1-042, and F1-048. F1-044 is not omitted by error. Its Packet-B RealizationLevel=Present code records an attributed participatory consequence—the authoritative water-line authorization and implementation sequence. Contrast E, however, uses the separately frozen project-level RealizationOutcome, under which Frontier is not classified in the conditioned/altered project-outcome cell. The two fields are distinct under the Packet-B separation rule (Appendix D, §1): consequence attribution is not automatically treated as project-path conditioning. This is not an attribution table: project conditioning may arise from factors unrelated to the observed participatory act. With only two silent cases, the comparison remains descriptive.

ContrastE = DescriptiveOnly_PrecisionGateNotMet