diff --git a/AUDIT.md b/AUDIT.md index 5992913..641a7db 100644 --- a/AUDIT.md +++ b/AUDIT.md @@ -79,14 +79,23 @@ The website remains an introductory projection, not a replacement for repository This cleanup makes evidence debt visible; it does not invent sources or claim that every exploratory argument is established. Several older analysis, ideology, and futures pages still need source-by-source research review. That work should happen in bounded evidence passes and should preserve the distinction between observation, interpretation, metaphor, and normative claim. -Two draft pull requests remain separate by design: - -- PR #6 develops the coherent-systems paper and should rebase or reconcile any shared navigation files before merge. -- PR #7 is the older isolated website editorial/accessibility prototype. The approved site now carries its central project-first and accessible-theme direction; do not merge the alternate review builder as a second website system. Preserve any still-useful review notes before closing it as superseded. - -Neither draft was merged into this audit branch, and neither was treated as current published state. - -Two pages in the separate `Root-Sequence/wiki` repository still link to the former `core/legible-systems.md` address on its current `main` branch. [Wiki PR #1](https://github.com/Root-Sequence/wiki/pull/1) is the prepared two-page update. Merge this cleanup first so the new canonical destination exists, then merge the Wiki PR immediately afterward. +Post-audit follow-up on 2026-09-18 retired both pre-cleanup pull requests +without merging their outdated structures: + +- PR #7 was closed after its useful project-first and accessibility direction + had been incorporated into the approved single-page website. +- PR #6 was closed after [PR #13](https://github.com/Root-Sequence/root-sequence/pull/13) + preserved its twelve-file coherent-systems paper package on a fresh branch + from current `main`, repaired retired paths, and reconciled its useful + cross-domain additions with their canonical pages. The paper remains + `v0.1-dev`, AI-assisted, author review pending, and not approved for release. + Rae approved the technical preservation and routing work on 2026-09-18; that + approval does not extend to the manuscript's claims or publication. + +The separate Wiki's former Legible Systems routes were corrected by merged +[Wiki PR #1](https://github.com/Root-Sequence/wiki/pull/1). Merged +[Wiki PR #2](https://github.com/Root-Sequence/wiki/pull/2) later added the +Root Sequence name and provenance entry. ## Verification record diff --git a/IDEA_TRAIL_GRAPH.yml b/IDEA_TRAIL_GRAPH.yml index 730d1d1..4841bcb 100644 --- a/IDEA_TRAIL_GRAPH.yml +++ b/IDEA_TRAIL_GRAPH.yml @@ -1,5 +1,5 @@ version: 0.3 -updated: 2026-09-16 +updated: 2026-09-18 canonical_trails: IDEA_TRAILS.md metadata_convention: IDEA_TRAIL_METADATA.md @@ -77,6 +77,16 @@ documents: - stewardship-authority - accountability-externalization + - repo: Root-Sequence/root-sequence + path: research/papers/coherent-systems/README.md + role: research + trails: + - intelligence-authority + - stewardship-authority + - accountability-externalization + - collective-judgment-dissent + - accessibility-participation + - repo: Root-Sequence/coherent-world path: no-one-noticed/research/INTELLIGENCE-ECOLOGY-AND-AURYN.md role: narrative diff --git a/IDEA_TRAIL_INDEX.md b/IDEA_TRAIL_INDEX.md index 29c2b8f..17afbb9 100644 --- a/IDEA_TRAIL_INDEX.md +++ b/IDEA_TRAIL_INDEX.md @@ -49,7 +49,7 @@ This index is generated from [`IDEA_TRAIL_GRAPH.yml`](IDEA_TRAIL_GRAPH.yml) and **Stable ID:** `stewardship-authority` -- **Research:** [`community-infrastructure/docs/research/karrot-hands-on.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/docs/research/karrot-hands-on.md) · [`root-sequence/analysis/collective-judgment-and-manufactured-consensus.md`](https://github.com/Root-Sequence/root-sequence/blob/main/analysis/collective-judgment-and-manufactured-consensus.md) · [`root-sequence/analysis/externalized-responsibility.md`](https://github.com/Root-Sequence/root-sequence/blob/main/analysis/externalized-responsibility.md) · [`root-sequence/concepts/intelligence-ecology.md`](https://github.com/Root-Sequence/root-sequence/blob/main/concepts/intelligence-ecology.md) +- **Research:** [`community-infrastructure/docs/research/karrot-hands-on.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/docs/research/karrot-hands-on.md) · [`root-sequence/analysis/collective-judgment-and-manufactured-consensus.md`](https://github.com/Root-Sequence/root-sequence/blob/main/analysis/collective-judgment-and-manufactured-consensus.md) · [`root-sequence/analysis/externalized-responsibility.md`](https://github.com/Root-Sequence/root-sequence/blob/main/analysis/externalized-responsibility.md) · [`root-sequence/concepts/intelligence-ecology.md`](https://github.com/Root-Sequence/root-sequence/blob/main/concepts/intelligence-ecology.md) · [`root-sequence/research/papers/coherent-systems/README.md`](https://github.com/Root-Sequence/root-sequence/blob/main/research/papers/coherent-systems/README.md) - **Human practice:** [`beinghumane-atlas/seeds/humane-collective-decision-architecture.md`](https://github.com/Root-Sequence/beinghumane-atlas/blob/main/seeds/humane-collective-decision-architecture.md) · [`beinghumane-guide/README.md`](https://github.com/Root-Sequence/beinghumane-guide/blob/main/README.md) · [`beinghumane-guide/notes/responsibility-accountability-and-consequences.md`](https://github.com/Root-Sequence/beinghumane-guide/blob/main/notes/responsibility-accountability-and-consequences.md) - **Embodied practice:** [`liberation-mass/README.md`](https://github.com/Root-Sequence/liberation-mass/blob/main/README.md) · [`liberation-mass/docs/roles.md`](https://github.com/Root-Sequence/liberation-mass/blob/main/docs/roles.md) - **Real-system work:** [`community-infrastructure/docs/concepts/collective-decision-making.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/docs/concepts/collective-decision-making.md) · [`community-infrastructure/docs/concepts/community-knowledge.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/docs/concepts/community-knowledge.md) · [`community-infrastructure/docs/concepts/groups-and-membership.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/docs/concepts/groups-and-membership.md) · [`community-infrastructure/docs/concepts/resource-commons.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/docs/concepts/resource-commons.md) · [`community-infrastructure/docs/design/threat-model.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/docs/design/threat-model.md) · [`community-infrastructure/docs/design/trust-and-safety.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/docs/design/trust-and-safety.md) @@ -93,7 +93,7 @@ This index is generated from [`IDEA_TRAIL_GRAPH.yml`](IDEA_TRAIL_GRAPH.yml) and **Stable ID:** `accessibility-participation` -- **Research:** [`community-infrastructure/docs/research/space-access-interoperability.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/docs/research/space-access-interoperability.md) · [`root-sequence/analysis/spatial-liberation.md`](https://github.com/Root-Sequence/root-sequence/blob/main/analysis/spatial-liberation.md) +- **Research:** [`community-infrastructure/docs/research/space-access-interoperability.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/docs/research/space-access-interoperability.md) · [`root-sequence/analysis/spatial-liberation.md`](https://github.com/Root-Sequence/root-sequence/blob/main/analysis/spatial-liberation.md) · [`root-sequence/research/papers/coherent-systems/README.md`](https://github.com/Root-Sequence/root-sequence/blob/main/research/papers/coherent-systems/README.md) - **Human practice:** [`beinghumane-guide/README.md`](https://github.com/Root-Sequence/beinghumane-guide/blob/main/README.md) - **Embodied practice:** [`liberation-mass/README.md`](https://github.com/Root-Sequence/liberation-mass/blob/main/README.md) · [`liberation-mass/docs/resilience-and-low-tech.md`](https://github.com/Root-Sequence/liberation-mass/blob/main/docs/resilience-and-low-tech.md) · [`liberation-mass/docs/safety/accessibility.md`](https://github.com/Root-Sequence/liberation-mass/blob/main/docs/safety/accessibility.md) · [`liberation-mass/docs/spaces/spaces.md`](https://github.com/Root-Sequence/liberation-mass/blob/main/docs/spaces/spaces.md) · [`liberation-mass/docs/spatial-principles.md`](https://github.com/Root-Sequence/liberation-mass/blob/main/docs/spatial-principles.md) - **Real-system work:** [`community-infrastructure/docs/concepts/community-assistant.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/docs/concepts/community-assistant.md) · [`community-infrastructure/docs/concepts/community-discovery.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/docs/concepts/community-discovery.md) · [`community-infrastructure/docs/concepts/community-home.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/docs/concepts/community-home.md) · [`community-infrastructure/docs/concepts/discovery-profile.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/docs/concepts/discovery-profile.md) · [`community-infrastructure/docs/concepts/space-mobility-and-access.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/docs/concepts/space-mobility-and-access.md) · [`community-infrastructure/docs/design/accessibility.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/docs/design/accessibility.md) ★ · [`community-infrastructure/docs/design/resilience-and-graceful-degradation.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/docs/design/resilience-and-graceful-degradation.md) · [`community-infrastructure/prototypes/community-assistant/README.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/prototypes/community-assistant/README.md) · [`community-infrastructure/prototypes/community-discovery/README.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/prototypes/community-discovery/README.md) · [`community-infrastructure/prototypes/degraded-mode/README.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/prototypes/degraded-mode/README.md) · [`community-infrastructure/prototypes/space-access/README.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/prototypes/space-access/README.md) @@ -149,7 +149,7 @@ This index is generated from [`IDEA_TRAIL_GRAPH.yml`](IDEA_TRAIL_GRAPH.yml) and **Stable ID:** `intelligence-authority` -- **Research:** [`community-infrastructure/docs/research/local-community-assistant.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/docs/research/local-community-assistant.md) · [`liberated-intelligence/README.md`](https://github.com/Root-Sequence/liberated-intelligence/blob/main/README.md) · [`root-sequence/concepts/intelligence-ecology.md`](https://github.com/Root-Sequence/root-sequence/blob/main/concepts/intelligence-ecology.md) +- **Research:** [`community-infrastructure/docs/research/local-community-assistant.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/docs/research/local-community-assistant.md) · [`liberated-intelligence/README.md`](https://github.com/Root-Sequence/liberated-intelligence/blob/main/README.md) · [`root-sequence/concepts/intelligence-ecology.md`](https://github.com/Root-Sequence/root-sequence/blob/main/concepts/intelligence-ecology.md) · [`root-sequence/research/papers/coherent-systems/README.md`](https://github.com/Root-Sequence/root-sequence/blob/main/research/papers/coherent-systems/README.md) - **Human practice:** [`beinghumane-atlas/seeds/intelligence-ecology-and-selection-environments.md`](https://github.com/Root-Sequence/beinghumane-atlas/blob/main/seeds/intelligence-ecology-and-selection-environments.md) · [`beinghumane-guide/README.md`](https://github.com/Root-Sequence/beinghumane-guide/blob/main/README.md) - **Real-system work:** [`coherent-computing/docs/conceptual-model.md`](https://github.com/Root-Sequence/coherent-computing/blob/main/docs/conceptual-model.md) · [`community-infrastructure/docs/concepts/community-assistant.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/docs/concepts/community-assistant.md) · [`community-infrastructure/docs/concepts/permissions-and-consent.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/docs/concepts/permissions-and-consent.md) · [`community-infrastructure/prototypes/community-assistant/README.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/prototypes/community-assistant/README.md) - **Speculative design:** [`coherent-world/README.md`](https://github.com/Root-Sequence/coherent-world/blob/main/README.md) · [`coherent-world/world/resilience-failure-and-continuity.md`](https://github.com/Root-Sequence/coherent-world/blob/main/world/resilience-failure-and-continuity.md) @@ -220,7 +220,7 @@ This index is generated from [`IDEA_TRAIL_GRAPH.yml`](IDEA_TRAIL_GRAPH.yml) and **Stable ID:** `accountability-externalization` -- **Research:** [`liberated-intelligence/README.md`](https://github.com/Root-Sequence/liberated-intelligence/blob/main/README.md) · [`root-sequence/analysis/collective-judgment-and-manufactured-consensus.md`](https://github.com/Root-Sequence/root-sequence/blob/main/analysis/collective-judgment-and-manufactured-consensus.md) · [`root-sequence/analysis/externalized-responsibility.md`](https://github.com/Root-Sequence/root-sequence/blob/main/analysis/externalized-responsibility.md) ★ · [`root-sequence/analysis/spatial-liberation.md`](https://github.com/Root-Sequence/root-sequence/blob/main/analysis/spatial-liberation.md) · [`root-sequence/concepts/intelligence-ecology.md`](https://github.com/Root-Sequence/root-sequence/blob/main/concepts/intelligence-ecology.md) · [`root-sequence/systems/resilience-and-graceful-degradation.md`](https://github.com/Root-Sequence/root-sequence/blob/main/systems/resilience-and-graceful-degradation.md) +- **Research:** [`liberated-intelligence/README.md`](https://github.com/Root-Sequence/liberated-intelligence/blob/main/README.md) · [`root-sequence/analysis/collective-judgment-and-manufactured-consensus.md`](https://github.com/Root-Sequence/root-sequence/blob/main/analysis/collective-judgment-and-manufactured-consensus.md) · [`root-sequence/analysis/externalized-responsibility.md`](https://github.com/Root-Sequence/root-sequence/blob/main/analysis/externalized-responsibility.md) ★ · [`root-sequence/analysis/spatial-liberation.md`](https://github.com/Root-Sequence/root-sequence/blob/main/analysis/spatial-liberation.md) · [`root-sequence/concepts/intelligence-ecology.md`](https://github.com/Root-Sequence/root-sequence/blob/main/concepts/intelligence-ecology.md) · [`root-sequence/research/papers/coherent-systems/README.md`](https://github.com/Root-Sequence/root-sequence/blob/main/research/papers/coherent-systems/README.md) · [`root-sequence/systems/resilience-and-graceful-degradation.md`](https://github.com/Root-Sequence/root-sequence/blob/main/systems/resilience-and-graceful-degradation.md) - **Human practice:** [`beinghumane-atlas/seeds/humane-collective-decision-architecture.md`](https://github.com/Root-Sequence/beinghumane-atlas/blob/main/seeds/humane-collective-decision-architecture.md) · [`beinghumane-atlas/seeds/intelligence-ecology-and-selection-environments.md`](https://github.com/Root-Sequence/beinghumane-atlas/blob/main/seeds/intelligence-ecology-and-selection-environments.md) · [`beinghumane-guide/README.md`](https://github.com/Root-Sequence/beinghumane-guide/blob/main/README.md) · [`beinghumane-guide/notes/responsibility-accountability-and-consequences.md`](https://github.com/Root-Sequence/beinghumane-guide/blob/main/notes/responsibility-accountability-and-consequences.md) ★ - **Embodied practice:** [`liberation-mass/docs/safety/consent-practices.md`](https://github.com/Root-Sequence/liberation-mass/blob/main/docs/safety/consent-practices.md) - **Real-system work:** [`coherent-computing/docs/conceptual-model.md`](https://github.com/Root-Sequence/coherent-computing/blob/main/docs/conceptual-model.md) · [`community-infrastructure/docs/concepts/collective-decision-making.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/docs/concepts/collective-decision-making.md) · [`community-infrastructure/docs/concepts/community-home.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/docs/concepts/community-home.md) · [`community-infrastructure/docs/concepts/needs-and-offers.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/docs/concepts/needs-and-offers.md) · [`community-infrastructure/docs/concepts/permissions-and-consent.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/docs/concepts/permissions-and-consent.md) · [`community-infrastructure/docs/design/threat-model.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/docs/design/threat-model.md) · [`community-infrastructure/docs/design/trust-and-safety.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/docs/design/trust-and-safety.md) · [`community-infrastructure/prototypes/needs-offers/README.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/prototypes/needs-offers/README.md) @@ -235,7 +235,7 @@ This index is generated from [`IDEA_TRAIL_GRAPH.yml`](IDEA_TRAIL_GRAPH.yml) and **Stable ID:** `collective-judgment-dissent` -- **Research:** [`root-sequence/analysis/collective-judgment-and-manufactured-consensus.md`](https://github.com/Root-Sequence/root-sequence/blob/main/analysis/collective-judgment-and-manufactured-consensus.md) ★ +- **Research:** [`root-sequence/analysis/collective-judgment-and-manufactured-consensus.md`](https://github.com/Root-Sequence/root-sequence/blob/main/analysis/collective-judgment-and-manufactured-consensus.md) ★ · [`root-sequence/research/papers/coherent-systems/README.md`](https://github.com/Root-Sequence/root-sequence/blob/main/research/papers/coherent-systems/README.md) - **Human practice:** [`beinghumane-atlas/seeds/humane-collective-decision-architecture.md`](https://github.com/Root-Sequence/beinghumane-atlas/blob/main/seeds/humane-collective-decision-architecture.md) - **Real-system work:** [`community-infrastructure/docs/concepts/collective-decision-making.md`](https://github.com/Root-Sequence/community-infrastructure/blob/main/docs/concepts/collective-decision-making.md) - **Speculative design:** [`coherent-world/README.md`](https://github.com/Root-Sequence/coherent-world/blob/main/README.md) · [`coherent-world/world/collective-decision-and-dissent.md`](https://github.com/Root-Sequence/coherent-world/blob/main/world/collective-decision-and-dissent.md) diff --git a/README.md b/README.md index 46102ee..421e345 100644 --- a/README.md +++ b/README.md @@ -48,6 +48,7 @@ The Liberation Cycle, Visibility Loop, and Roots → Signals → Growth → Frui ## 📂 Repository map - [`site/`](site/README.md) — reader-facing Public Seed website source; not a replacement for research or the Wiki +- [`research/`](research/README.md) — versioned synthesis papers; underlying concepts and evidence retain their canonical homes - [`founding-texts/`](founding-texts/README.md) — founding texts, early models, manifestos, normative essays, and historical orientations - [`systems/`](systems/README.md) — analytical principles, models, and methods for understanding system behavior - [`concepts/`](concepts/README.md) — shared definitions, cross-project frameworks, design principles, and methods @@ -69,6 +70,7 @@ The Liberation Cycle, Visibility Loop, and Roots → Signals → Growth → Frui Suggested entry points: - [Public Seed website source](site/README.md) — approved single-page introduction, publication workflow, and deployment records +- [Coherence Is a Systems Property](research/papers/coherent-systems/README.md) — v0.1-dev living-paper scaffold; AI-assisted and awaiting author review - [Ecosystem Map](ECOSYSTEM.md) — understand the wider project family - [Root Sequence Wiki](https://github.com/Root-Sequence/wiki) — browse terms, phrases, projects, provenance, and conceptual archaeology - [Idea Trails](IDEA_TRAILS.md) — understand the recurring questions and how each project transforms them diff --git a/THOUGHT_ROUTING.md b/THOUGHT_ROUTING.md index 4c7f751..85a5e77 100644 --- a/THOUGHT_ROUTING.md +++ b/THOUGHT_ROUTING.md @@ -45,6 +45,16 @@ Check, as relevant: Search for the mechanism and question, not only the exact phrase. Similar wording may hide different ideas; different wording may describe the same one. +#### Build on existing work; name the actual gap + +**Clarified by the author, 2026-09-16:** Interest in an idea or outside guidance is not a request to create another framework. Look for relevant established work as well as existing project documents. Reuse what already addresses the need, connect missing references, and extend only what remains insufficiently covered. A useful synthesis need not claim novelty. + +Before proposing a new framework, taxonomy, repository, or tracking layer, state what the existing options cannot do and ask whether a new structure is wanted. Do not make new terminology the default response to enthusiasm. A passage in an existing home may be enough. + +During a routing detour, preserve the active project's next unresolved question in its existing record. After the update, return a short recap of what we were doing and that question, rather than another expanding plan. Keep proposed interpretations separate from author decisions. A request to save insights does not answer a previously open conceptual question. + +Apply the [participation and review-burden lesson](concepts/legible-systems.md#participation-sensemaking-and-review-burden): reduce reconstruction work, do not count document production as comprehension, and check unresolved dependencies before consequential commitments. Exploratory ideas may remain separate or be dropped; the map does not have to make every connection fit. + ### 3. Classify on separate axes Do not collapse these into one vague status. diff --git a/concepts/legible-systems.md b/concepts/legible-systems.md index acb3b44..54e6e57 100644 --- a/concepts/legible-systems.md +++ b/concepts/legible-systems.md @@ -6,6 +6,8 @@ **Evidence boundary:** This page proposes design guidance. Each translation into technical, institutional, educational, or physical domains needs independent validation.
**First formalized here:** 2026-09-06
+**Recent addition:** [Participation, sensemaking, and review burden](#participation-sensemaking-and-review-burden) connects the existing principle to checked W3C guidance and records practical lessons from the living-paper work. + ## Core idea A **legible system** can be used simply without requiring its workings to remain mysterious. @@ -343,3 +345,47 @@ More generally: > **A humane system should help the people inside it understand the system, participate in it, question it, modify it, and eventually need its designers less.** That is the direction of travel, not a claim that every system can or should become equally simple. + +--- + +## Participation, sensemaking, and review burden + +**Added:** 2026-09-16, from the living-paper conversation. **Status:** working design interpretation; no usability or efficacy result. This extends the existing treatment, not a new framework. Earlier material above is preserved. + +> **Understanding can develop through supported participation; it needn't arrive fully formed before participation begins.** + +This is a project formulation, not a quotation from W3C or a claim of scholarly novelty. Invite questions, examples, corrections, and uncertainty before demanding a finished explanation. Keep relevant evidence available. Support for exploration does not authorize consequential action without adequate understanding, consent, or safeguards. + +### Sensemaking without forced agreement + +Weick, Sutcliffe, and Obstfeld's *Organizing and the Process of Sensemaking* (2005) connects explicit interpretations of circumstances with action. The publisher's abstract and bibliographic details were checked; this is not a full-text review.[^sensemaking2005] + +Our application distinguishes understanding what someone says, understanding why a choice seemed reasonable, and concluding that an account fits the evidence. One does not establish the others. Assistance should help make an interpretation inspectable and revisable, not manufacture agreement or mistake fluency for truth. “Supported sensemaking” is descriptive language here, not a named replacement for existing research. + +### Do not transfer the burden and call it assistance + +A proposed failure pattern in AI-assisted work is: more documents produced, more apparent progress, more unacknowledged reading and verification work for the person. This is a workflow hypothesis, not a measured result from this project. Review effort, comprehension, unresolved decisions, and the ability to resume deserve attention alongside output volume. + +Keep the current passage, source locator, decision status, and next question together. Provide reminders with actionable alternatives or a route to someone able to change the constraint. A warning alone cannot create time, resources, or decision authority. Explanations must remain challengeable, and neither refusal nor uncertainty should count as failure to understand. + +### Existing guidance to use before adding new rules + +The following sources were checked on 2026-09-16 at the stated scope. Their roles differ; none certifies Root Sequence, a whole framework, or an implementation. Extending web guidance into institutional or collaborative settings is our proposed application, not demonstrated universal transfer. + +| Source and exact edition | Material inspected and possible use | Boundary | +|---|---|---| +| [Making Content Usable for People with Cognitive and Learning Disabilities](https://www.w3.org/TR/2021/NOTE-coga-usable-20210429/), Working Group Note, 29 April 2021 | Abstract/status, easy-language summary, objective/pattern headings, and focus/recovery passages: familiar steps, less reliance on memory, summaries, error recovery, help, and involving affected users. | Supplemental guidance, not additional WCAG conformance requirements. No pattern-by-pattern audit or user testing performed here. | +| [Web Platform Design Principles](https://www.w3.org/TR/2026/NOTE-design-principles-20260914/), Group Note, 14 September 2026 | Abstract/status and section 1.1: prioritize user needs while acknowledging implementation dependencies and finite resources. | API-design guidance, not proof of a universal priority rule or endorsement of our project. | +| [Privacy Principles](https://www.w3.org/TR/2025/STMT-privacy-principles-20250515/), W3C Statement, 15 May 2025 | Status and section 1.1.2: privacy labor concerns offloading protective work onto the people affected by data processing. | A grounded comparison for responsibility transfer; not a reason to remove meaningful choice or consent. | +| [Ethical Web Principles](https://www.w3.org/TR/2024/STMT-ethical-web-principles-20241212/), W3C Statement, 12 December 2024 | Sections 2.7–2.11: verifiable information, individual control, reduced centralization, environmental effects, inspectability, and interoperable alternatives. | Normative design principles, not evidence that actual web systems already deliver them. | +| [Web Sustainability Guidelines](https://www.w3.org/TR/2026/DNOTE-web-sustainability-guidelines-20260820/), Group Note Draft, 20 August 2026 | Abstract/status and section headings: impacts on people and the environment, infrastructure, product management, maintenance, and resource budgets. | Initial orientation only; not a completed standard, impact calculation, or full guideline review. | + +W3C's cognitive guidance explicitly goes beyond WCAG's requirements. The practical lesson for this project is to examine unmet needs even when formal requirements have been met; it is not permission to disregard those requirements. + +### Learning while building + +Exploration can branch before every question is resolved. Before a consequential commitment, identify what it depends on, which assumptions remain unsettled, and who bears failure. Retain reversible trials and revision paths where possible. Do not force every idea into a final harmonious account: some connections may fail, remain unresolved, or be better kept separate. + +Use the existing [thought-routing convention](../THOUGHT_ROUTING.md) to record these distinctions without another tracking system. The [paper's drafting instructions](../research/papers/coherent-systems/AGENTS.md) apply them to passage-level author review. These links describe saved process guidance, not an implemented or validated interface. + +[^sensemaking2005]: Karl E. Weick, Kathleen M. Sutcliffe, and David Obstfeld (2005), *Organizing and the Process of Sensemaking*, Organization Science 16(4), 409–421. DOI: 10.1287/orsc.1050.0133. [Publisher abstract and citation](https://pubsonline.informs.org/doi/10.1287/orsc.1050.0133). No claim that an intelligible interpretation is therefore correct. diff --git a/founding-texts/universal-coherence-framework.md b/founding-texts/universal-coherence-framework.md index c40bc71..3ef61bf 100644 --- a/founding-texts/universal-coherence-framework.md +++ b/founding-texts/universal-coherence-framework.md @@ -6,6 +6,8 @@ This page records an earlier compact Root Sequence treatment. The dedicated UCF repository owns the current framework, models, tests, and project-specific terminology. For the smaller systems-layer treatment of coherence inside this repository, see [`../systems/coherence/`](../systems/coherence/README.md). +The six headings below are a local explanatory sequence, not UCF's four analytical layers or its optional state labels. Statements about natural drift are working hypotheses, not demonstrated laws. Functional fit does not by itself establish wider benefit or legitimate authority. The developing paper [*Coherence Is a Systems Property*](../research/papers/coherent-systems/notes/ucf-reconciliation.md) records how it uses and limits the UCF material; both remain subject to author review. + The Universal Coherence Framework (UCF) describes how systems align—or fail to align—with reality over time. It is not a fixed model. It is a way of observing whether a system holds together internally, interacts accurately with the world, and sustains itself without generating hidden instability. diff --git a/research/README.md b/research/README.md new file mode 100644 index 0000000..5f28b10 --- /dev/null +++ b/research/README.md @@ -0,0 +1,15 @@ +# Research syntheses + +This directory holds versioned papers that synthesize Root Sequence's existing research. It does not replace the canonical treatments in `concepts/`, `systems/`, `analysis/`, or `futures/`. + +## Living papers + +- [Coherence Is a Systems Property](papers/coherent-systems/README.md) — **v0.1-dev; AI-assisted working draft; author review pending**. Charter, annotated argument, claims ledger, source map, and research questions. Not a reviewed release or a validated theory. + +A paper proposes a synthesis; the underlying research can contradict and revise it. Do not move material into this directory merely to make the paper appear comprehensive. + +## Publication boundary + +Public branches and pull requests in this repository are public. They are not private review spaces. Only public-safe synthesis belongs here; private notes and unpublished fiction retain their existing homes and permissions. Website presentation is separate and requires an explicit publication pass. + +[Repository map](../root_map.md) · [Contribution guidance](../contributing.md) diff --git a/research/papers/coherent-systems/AGENTS.md b/research/papers/coherent-systems/AGENTS.md new file mode 100644 index 0000000..28462b5 --- /dev/null +++ b/research/papers/coherent-systems/AGENTS.md @@ -0,0 +1,33 @@ +# Drafting and author review + +**Requested workflow, recorded 2026-09-16.** Applies to the paper and its supporting notes. This process does not approve any manuscript claim or authorize publication. + +## Work through questions + +Retrieve relevant existing project material before asking the author to repeat it. Draft one manageable passage or decision at a time. When an ambiguity changes meaning, scope, claims, values, attribution, or publication, ask before making that decision on the author's behalf. + +Default to one substantive question at a time. Supply the short passage, example, or source locator needed to answer. Two or three concrete alternatives can help, but leave room for a mixture, neither, uncertainty, or different wording. Avoid leading questions that make one answer sound like the only reasonable choice. Routine formatting and other reversible editorial maintenance do not need repeated permission. + +Accept fragments, examples, corrections, and reactions as input. Label your proposed synthesis as interpretation; check consequential changes rather than treating your formulation as the author's own words. Show what changed and why after an answer. Do not postpone all author input until a large finished draft exists. + +## Support review without replacing judgment + +Keep the relevant text and source material available during review. Do not require an unaided explanation from memory or an improvised defense as proof of participation. Help develop a written explanation that the author can inspect, correct, qualify, or reject. Pause or restate a question when needed. + +Source checks and independent evaluation remain necessary where the claim requires them. Do not confuse conceptual agreement with source verification, approval of wording, or approval to publish. A request to continue, silence, or approval of one passage does not approve the entire manuscript. No personal health disclosure is required to use this workflow. + +## Keep the record usable + +Maintain a brief decision note in the existing claim, source, or change record: author-stated intent, assistant interpretation, affected passage, unresolved question, and review status. Record only decisions actually made. Do not create a new tracking system or ask the author to maintain duplicate logs. + +Keep earlier meanings, counterarguments, and source visibility intact. Do not invent earlier contributions, quotations, personal experience, or approvals. Do not publish private conversation excerpts or personal details merely because they explain the process. + +Use [AI-ASSISTANCE.md](AI-ASSISTANCE.md) for disclosure and review distinctions and [README.md](README.md) for paper scope. These repository instructions are not a claim that every future tool or conversation automatically loads them. + +## Extend what exists and return to the active question + +**Author clarification, 2026-09-16:** The aim is to address gaps in existing work, not generate new frameworks by default. Check established literature and existing project treatments, identify the specific missing piece, and add only that piece. Preserve the option that the useful contribution is a synthesis or application, not a novel method. An enthusiastic response to a term is not authorization to create a taxonomy or new project. + +Apply [the existing Legible Systems treatment](../../../concepts/legible-systems.md#participation-sensemaking-and-review-burden) to this collaboration: more prose can transfer more review work to the author. Keep enough context for the next decision, avoid expanding unreviewed dependent claims, and provide a short re-entry recap after a research or routing detour. Do not put the burden of restoring context back on the author. + +Use [THOUGHT_ROUTING.md](../../../THOUGHT_ROUTING.md) for cross-project routing. Keep open questions in the existing claim record. Neither helpful guidance nor a readable explanation replaces evidence, independent review where needed, or consent to consequential actions. diff --git a/research/papers/coherent-systems/AI-ASSISTANCE.md b/research/papers/coherent-systems/AI-ASSISTANCE.md new file mode 100644 index 0000000..9f8f103 --- /dev/null +++ b/research/papers/coherent-systems/AI-ASSISTANCE.md @@ -0,0 +1,78 @@ +# AI assistance, verification, and author review + +**Record started:** 2026-09-16. **Applies to:** the v0.1-dev living paper and its documented UCF connection. **Status:** working disclosure, not an author-approved submission statement. The personal explanation below was specifically requested for inclusion; that limited authorization does not approve the manuscript as a whole. + +## Why I use AI: accessibility and developing the work + +**Author-requested explanation, 2026-09-16:** The author clarified that accessibility and developing ideas are both purposes of the assistance, responded positively to the combined wording, and explicitly requested that it be added in the continuing booking-service discussion. This section preserves that wording without adding a diagnosis or claiming completed source review. + +> **I use AI both as an accessibility aid and as a tool for developing my work.** My ideas often emerge through questions, fragments, and conversations across many connected subjects. AI helps me gather that material, organize it, explore relationships, compare ideas with existing research, and develop drafts I can work through in manageable pieces. +> +> My contributions include the motivating questions, conceptual direction, examples, constraints, objections, and decisions that shape the work. The AI assistance includes substantive drafting and critique, not only editing. Working through questions and revisions helps me remain involved throughout, rather than having to organize or explain everything at once. +> +> That assistance does not make AI-generated claims evidence, and a draft does not automatically represent my endorsement. Sources, reasoning, and conclusions still need appropriate checking and review. + +## Disclosure suitable for the current draft + +This draft was developed with ChatGPT assistance for organizing project material, locating literature, drafting and critiquing arguments, and developing the synthetic example with associated Python code and tests. AI output is not treated as evidence. Source inspection, software checks, author review, and independent evaluation are separate activities. Source-by-source author review, approval of the current manuscript, and independent evaluation remain incomplete. + +The author has contributed motivating questions, project direction, connections among projects, constraints, objections, and revision decisions through conversation. These contributions are already part of developing the work; final manuscript approval is a different activity. Exact quotations or attribution to earlier conversations require traceable sources and appropriate permission. Neither this description nor a request to continue implies approval of every synthesized claim or verification of all references. + +This assistance was more extensive than grammar correction. The record should not be narrowed to “language editing” to make it appear more acceptable. It also should not claim that the author has personally verified every reference or approved every argument when that review has not occurred. + +## What assistance has covered + +| Activity | Actual role | Limit and required review | +|---|---|---| +| Project synthesis | Organizing public project concepts, comparing definitions, proposing structure and claims | User authorization to develop a draft is not endorsement of every resulting proposition. | +| Literature work | Searching, opening primary sources, inspecting selected text and PDF figures, drafting comparisons | Retrieval and AI-assisted inspection are not independent human reading, exhaustive search, or validation of the cited studies. | +| Writing and critique | Drafting substantial prose, proposing examples, objections, and evidence classifications | Agreement between AI outputs is not independent peer review. The author must decide what to retain. | +| Computational work | Earlier assistance produced the UCF booking model and its implementation checks | A code test checks specified behavior. It does not validate a scientific framework, study finding, or allocation rule. | +| Repository work | Preparing files, references, cross-links, change records, and draft-branch updates | A saved commit is not approval, deployment, or a reviewed publication. | + +The external-evidence pass added no patient data and ran no empirical or participant study. It did not rerun or replicate the Obermeyer analysis. The earlier booking checks remain documented in their own revision and recovery records rather than being relabeled as tests performed in that pass. + +## Tool record and reproducibility limits + +The platform used is ChatGPT, with web retrieval, GitHub actions, and local Python where needed. Exact model identifiers and settings for every earlier drafting session are not preserved in the repository; they must not be reconstructed from assumptions about a product name. No claim of full conversational reproducibility is made. + +For future sessions, record the tool and visible model/version when available, date, task, affected paths, code inputs and outputs, source locators, and human corrections. If a field is unavailable, say so. Preserve relevant public-safe task instructions, not hidden reasoning or unrelated private conversations. The versioned text and runnable code are the durable artifacts. + +The source map records what was actually inspected. Validation records must distinguish a retrieved document, verified bibliographic metadata, checked passages, reproduced arithmetic, and independent replication. None should stand in for another. + +## Question-led author review + +The author requested clarification questions throughout drafting rather than a large document followed by an all-at-once defense. [AGENTS.md](AGENTS.md) records that working method. Present one small passage or example, provide the relevant context, and ask one meaningful question. Responses may be fragments, examples, corrections, or uncertainty. The assistant proposes a synthesis and shows the resulting change; it must not silently decide unresolved meaning. + +Review may use written notes, source passages, repeated explanations, and pauses. It does not require unaided recall or fluent speech on demand. The standard is an inspectable account of the claim, its support, and the author's decisions, not a performance test. This support does not turn unchecked evidence into checked evidence. + +The checkpoints below are a sequence to work through with support, not a batch of homework. None is pre-approved: + +- [ ] Present a short part of section 4, an example, and the relevant source passage. Ask what is accurate, too strong, unclear, or missing; revise C-003 with the author's response. +- [ ] Present the healthcare case one finding at a time with its source locator. Work through what a reported quantity measures, recording any uncertainty and which source material the author has reviewed. +- [ ] Review C-001, C-003, and C-012 individually. Record accept, qualify, reject, or not yet decided for the exact passage and revision; do not infer reasons the author did not give. +- [ ] Review the abstract and disclosure in small units. Confirm scope, assistance, and remaining uncertainty separately from any eventual release or byline decision. + +The first checkpoint is now in progress: the author's booking-service response is recorded under C-003 in [claims.md](claims.md), with the safeguard proposal under C-006. It does not yet settle the definition or approve the section. + +Do not add an approved personal byline, institutional affiliation, peer-review claim, or release identifier before the appropriate decision is made. Project provenance is not a substitute for an eventual accountable author record. + +## Publication and privacy boundary + +Elsevier's author guidance, checked in the external-evidence pass on 2026-09-16, permits specified AI support but requires human oversight, verification, and disclosure; research-method uses require methodological documentation. It does not guarantee acceptance elsewhere or certify this draft's compliance.[^elsevier] + +Before submission, check the chosen venue's current requirements, actual tool terms, permissions, and data handling. Those terms and account settings have not been audited here. Neither a paid account nor a public repository proves that all confidentiality obligations are satisfied. + +This GitHub branch is public. The author-requested personal explanation is the public wording in the section above, not permission to disclose diagnoses, unrelated private conversations, patient records, or confidential manuscripts. Additional personal accounts or new publication destinations require the relevant permission. The package contains original notes and links, not copies of the cited papers or their figures. + +## Changes recorded 2026-09-16 + +Added the requested combined explanation of AI as both an accessibility aid and a tool for developing the work. Kept the substantive-assistance disclosure and verification requirements. Recorded that passage-level review has started, without marking manuscript approval or source-review checkboxes complete. + +The earlier process edit added the requested question-led workflow and clarified that conceptual contributions already made are distinct from incomplete review of the current manuscript. It replaced the checkpoint requiring an unaided explanation with supported, passage-level review. No research result, byline approval, or release status follows from either process edit. + +## References + +[^elsevier]: Elsevier, *Generative AI policies for journals*, “For authors,” disclosure, research-process/code-development FAQs, and documentation guidance. [Policy](https://www.elsevier.com/about/policies-and-standards/generative-ai-policies-for-journals), accessed in the earlier external-evidence pass on 2026-09-16; not rechecked for this process edit. A publisher's policy, not empirical support for the paper's claims. + +[Source map](notes/source-map.md) · [Claims](claims.md) · [Changelog](CHANGELOG.md) diff --git a/research/papers/coherent-systems/CHANGELOG.md b/research/papers/coherent-systems/CHANGELOG.md new file mode 100644 index 0000000..3246d31 --- /dev/null +++ b/research/papers/coherent-systems/CHANGELOG.md @@ -0,0 +1,93 @@ +# Changelog + +## v0.1-dev — 2026-09-18 — Repository-structure rescue + +Transferred the twelve-file paper package from development revision +`5c15f7fecd96924b9a18f6ff49d2cb01f2568dd0` onto the reorganized repository's +current `main` history. Repaired the Legible Systems references to use the +canonical `concepts/legible-systems.md` path and added current navigation from +the research index, repository front door, repository map, and Idea Trails. + +The earlier additions to Thought Routing, Legible Systems, and the historical +UCF overview were reconciled with their current canonical files instead of +restoring the retired `core/` structure. No manuscript claim, source-reading +record, author decision, byline, paper version, test status, or publication +status changed in this rescue. The review still resumes at section 4 / C-003, +and T-003 remains unrun. + +Rae Lovejoy approved this repository-structure rescue on 2026-09-18. That +approval covers preservation, placement, repaired routes, and navigation. It +does not approve the manuscript's claims, sources, byline, release status, or +publication. + +## v0.1-dev — 2026-09-16 — Existing-work gap filling and supported participation + +Recorded the author's clarification in `../../../THOUGHT_ROUTING.md` and paper-scoped `AGENTS.md`: use established work and existing project homes, name the actual gap, and do not turn an interesting term into a new framework by default. + +Extended `../../../concepts/legible-systems.md` with the conversation's working principle that understanding can develop through supported participation; the distinction between an intelligible account and an evidence-supported one; the risk of transferring review burden through more output; and checking dependencies before consequential commitments. These are design proposals and process lessons, not new empirical findings. + +The source/application table in that existing concept now records five W3C documents with their edition, status, inspected scope, and limits, plus Weick, Sutcliffe, and Obstfeld's sensemaking article at publisher-abstract coverage. W3C material beyond cognitive accessibility includes user priorities, privacy labor, control, inspectability, interoperability, and sustainability. Guidance and Statements are not treated as implementation validation or interchangeable with a conformance standard. Broader uses remain proposed applications. + +The human-scale translation belongs in the existing private BHIG `map/legible-systems.md`, not a public copy of private source. No new framework, taxonomy, repository, or document is required by this pass. Existing bodies are retained with additions. The previously requested AI-use explanation remains in `AI-ASSISTANCE.md` and is not replaced. + +**Resume point:** section 4 / C-003, with C-006's safeguard discussion. The author has distinguished understandable situated reasoning from actual fit and target compliance, and asked for consequences to be visible without blaming people who lack information, resources, or authority. The remaining question is what would make a service's response coherent when capacity remains insufficient, and where the unresolved shortfall belongs in that assessment. No answer, new definition, or author approval is inferred from this routing request. T-003 remains unrun; the manuscript and UCF model are unchanged in this pass. + +## v0.1-dev — 2026-09-16 — Author input on situated reasoning and safeguards + +Added the requested combined first-person explanation of AI as an accessibility aid and a development tool to `AI-ASSISTANCE.md`, without adding a diagnosis. This scoped inclusion does not approve the manuscript, its byline, or publication elsewhere. The preceding question-led collaboration edit is recorded in that file and in the paper-scoped `AGENTS.md`. + +Recorded the author's booking-service response under C-003 and the safeguard direction under C-006. Kept author-stated intent, assistant synthesis, proposed interface text, and unresolved questions distinct. Preserved “malconsidered” as a tentative candidate rather than imposing it as technical vocabulary. Added Dekker's 2002 contextual-reconstruction paper at abstract-level coverage and documented a focused STPA reading on beliefs, actual conditions, and feedback. There are now nine bibliography entries; the new entry supports the ledger discussion, not an additional empirical finding in the manuscript. + +The original twelve propositions remain in place. Section 4 and the UCF model are not rewritten while the distinction between situated reasoning, target compliance, and coherence is still being clarified. No source-review or manuscript-approval checkbox was marked complete. No safeguard was implemented or evaluated, and no code, site, or empirical test was changed or run. + +## v0.1-dev — 2026-09-16 — External evidence (third pass) + +Added `notes/related-work.md`, `case-studies/healthcare-cost-proxy.md`, and `AI-ASSISTANCE.md`. Added three research references (STPA, Selbst and colleagues, and Obermeyer and colleagues), giving eight bibliography entries: seven external research/method sources and one internal UCF source. The separate publisher-policy citation is editorial context, not evidence for research claims. + +Revised C-001's provenance, C-003's distinction between interacting questions rather than independent stages, C-005's case support, and C-012's untested added-value claim. Expanded manuscript links and citations without removing the twelve-section structure, synthetic example, limitations, or UCF provenance. The abstract now describes the external case without claiming new field findings. + +Recorded exact source-reading limits, including targeted full-text passages beyond abstracts. Preserved the earlier source-check table as historical coverage, not a current claim that no further reading occurred. Added candidate comparison design and author checkpoints; T-003 remains unrun. No new participant research, patient data, independent replication, reviewer approval, or venue compliance claim. + +The disclosure accurately includes substantial drafting, comparison, research organization, and earlier code assistance, not merely grammar editing. Personal byline and substantive author review remain unresolved. Earlier versions remain in Git; no UCF source, site, main branch, private material, or fictional canon is changed by this pass. + + +## v0.1-dev — 2026-09-16 — UCF reconciliation and interruption recovery + +Added the dedicated UCF model and glossary to the source map; the initial scaffold had omitted them. Compared UCF's definitions with C-003 and changed “operational definition” to “working definition” until observations and rules are specified. Kept the four analytical layers distinct from the optional four-state taxonomy. + +Expanded section 2 with the implemented synthetic booking example and its boundary conditions. Changed allocation redistributes access without eliminating unmet work in the base fixture. Canonical code stays in UCF. Updated C-001, C-003, C-012, research-task statuses, the charter, bibliography, and navigation. Clarified that the one-period allocation example does not test C-002's claim about behavior persisting over time. + +Recovered the staged manuscript, charter, bibliography, and UCF note after interruption, then completed the companion ledger and status changes before updating the existing draft PR. Reran the eight implementation tests and 24-fixture report from source matching the saved Git blob hashes. No reproducible 700-case comparison was recovered; its progress-text claim is excluded from the evidence record. + +T-001 has an editorial case set, not independent reviewers. T-002 has code and passing implementation tests. T-003 remains unrun. No field results, framework validation, author approval, or deployment is claimed. Added a scope/provenance notice to Root Sequence's older UCF overview without replacing its body. + +The entries below describe the earlier scaffold as it was then; their source counts and unexecuted-test status are historical, not the current revision. + +## v0.1-dev — 2026-09-16 — Initial scaffold + +### Added + +- A bounded paper charter and annotated twelve-section argument with a provisional abstract. +- Twelve provisional claims, objections, support limits, evidence needs, and revision triggers. +- Ten open questions and three proposed tests; no test is represented as executed. +- Four bibliographic entries with source-review limits and a public-source provenance map. +- Research navigation and pointers from the root README and repository map. + +### Clarified from the planning outline + +- “From optimization to coherence” becomes **optimization within explicit boundaries**. The draft does not reject optimization as a technique. +- Functional coherence is distinguished from normative legitimacy, rather than silently defining coherence as goodness. +- Directional change does not mean inevitable or monotonic progress; no theorem about the impossibility of a final coherent state is asserted. +- Local-versus-systemic alignment is acknowledged as shared prior work, not claimed as an original discovery. +- Intelligence capability, moral direction, legitimate authority, and actual benefits remain separate questions. +- A working draft is not a reviewed release; AI-assisted prose and authorship require approval. + +### Boundaries + +Only public-safe synthesis and verified public repository sources were used for the research scaffold. No private source documents or unpublished fictional passages were imported. No story canon, website source, deployment, other repository, or external institution's work is modified by this change. + +No empirical findings, automatic update system, DOI, public-facing paper page, or complete literature review has been produced. Public repository review branches are public, not private staging. + +### Next revision + +Stress-test the definition and the service-access counterexample; complete the prior-work comparison; revise claims and abstract before treating v0.1 as a reviewed release. diff --git a/research/papers/coherent-systems/README.md b/research/papers/coherent-systems/README.md new file mode 100644 index 0000000..c451be2 --- /dev/null +++ b/research/papers/coherent-systems/README.md @@ -0,0 +1,81 @@ +# Coherence Is a Systems Property + +## Living-paper charter + +**Working subtitle:** Intelligence, Institutions, and the Conditions They Create +**Project:** Root Sequence +**Version:** v0.1-dev +**Started:** 2026-09-16 +**Status:** AI-assisted research scaffold; author review pending; not peer reviewed +**Release status:** No reviewed v0.1 release, DOI, or website publication is created by this scaffold. + +> Root Sequence tests, complicates, and revises the paper—not vice versa. + +### Purpose and audience + +Develop a bounded, examinable argument about how intelligent systems interact with their surrounding institutions, infrastructure, material environments, and affected people. The audience includes researchers, designers, builders, and technically literate general readers. + +The paper is a synthesis within Root Sequence, not the charter of a new institution or a replacement for the ecosystem's independent projects. + +### The question for v0.1 + +**What must be examined beyond the objectives of an individual agent to assess whether a change improves the conditions of the wider system?** + +The proposed contribution is a review method connecting system boundaries, assumptions, consequences, authority, material constraints, and revision over time. It now explicitly draws on UCF's four-layer model without requiring its four-state vocabulary or assuming universality. Whether that method adds anything to existing sociotechnical alignment and systems practice is an open research question, not a novelty claim. + +### Scope and non-goals + +v0.1 should define its terms, locate prior work, develop one counterexample to local-to-systemic inference, specify a small assessment procedure, and propose tests. + +It will not claim to solve society, predict AGI, prove inevitable progress, provide a universal coherence score, replace technical AI safety, or establish one final social arrangement. Intelligence is neither presumed benevolent nor presumed malevolent. Removing a monetary constraint is not the same as removing material, ecological, coordination, maintenance, or consent constraints. + +### Start reading + +1. [Argument and provisional abstract](paper.md). +2. [Claims ledger](claims.md), including objections and evidence gaps. +3. [Open questions and first research tests](open-questions.md). +4. [Source and provenance map](notes/source-map.md). +5. [Bibliography](references.bib) and [revision history](CHANGELOG.md). +6. [UCF reconciliation and worked-example status](notes/ucf-reconciliation.md). +7. [Related work](notes/related-work.md) and the [published healthcare case](case-studies/healthcare-cost-proxy.md). +8. [AI-assistance disclosure and author-review checkpoint](AI-ASSISTANCE.md). + +### What “living” means + +Substantive changes pass through review, update the claims ledger and changelog, and retain contrary evidence. Git history records revisions; it does not supply scholarly validation. This scaffold does not configure automatic rewriting, background research, or deployment. + +Development versions use `v0.1-dev`. A reviewed release should identify its date and exact commit. A later citation should identify the title, approved authors, version, date, and immutable revision. No DOI or author endorsement should be invented. + +### Evidence and editorial rules + +- Separate definitions, literature-supported arguments, inferences, hypotheses, and design commitments. A normative commitment is not an empirical discovery. +- Cite external factual claims with Markdown footnotes; record source coverage and version in the source map and bibliography. Abstract review is not full-text review. +- A counterexample can defeat a universal guarantee without establishing how common the failure is. Scenarios and fiction generate questions, not empirical validation. +- Keep dissent, privacy, legitimate authority, and real constraints visible. Coherence must not become a synonym for obedience or whatever the author prefers. +- Do not copy private notes, personal data, unpublished fiction, or story canon into this public directory. AI-assisted wording remains subject to author revision; the author list is unresolved until approved. + +### Canonical placement + +Canonical source for this paper: `Root-Sequence/root-sequence/research/papers/coherent-systems/`. + +Its conceptual inputs remain canonical in their existing folders. Coherent World remains a systems-design space; *No One Noticed* remains a narrative space, not evidence that a proposed transition will work. Cross-project context is routed through the existing [Ecosystem Map](../../../ECOSYSTEM.md), not a new atlas or repository. + +A future reader-facing rendering may live at `site/research/coherent-systems/`. That path is **proposed, not created here**; any rendering must derive from the canonical source and pass publication review. + +### External evidence and review status + +A targeted prior-work comparison and one published case have been added. They constrain C-001, C-003, and C-012 without turning internal project references into independent evidence. STPA is a candidate comparison method, not a baseline already evaluated. Author approval and the wider literature review remain pending. + +### Current exercise status + +The UCF booking example is implemented and has eight passing code/arithmetic checks. The four definition cases have an AI-assisted editorial treatment, not independent reviewers. The baseline comparison remains unrun. No field findings or added-value result is claimed. UCF owns the detailed model and code; the paper owns the synthesis. + +### Gate for the first reviewed release + +- Author approves the definition, scope, abstract, and byline. +- Central claims have appropriate evidence or explicit argument status; literature gaps remain visible. +- At least one worked example is analyzed against a baseline and a serious counterexample. +- Prior-work comparison describes overlap without unsupported claims of originality or superiority. +- Links, citations, visibility, and changelog are checked; any empirical tests actually performed are distinguished from proposals. + +[Research index](../../README.md) · [Root Sequence](../../../README.md) diff --git a/research/papers/coherent-systems/case-studies/healthcare-cost-proxy.md b/research/papers/coherent-systems/case-studies/healthcare-cost-proxy.md new file mode 100644 index 0000000..268e169 --- /dev/null +++ b/research/papers/coherent-systems/case-studies/healthcare-cost-proxy.md @@ -0,0 +1,38 @@ +# Healthcare cost as a proxy: a published case + +**Source:** Ziad Obermeyer, Brian Powers, Christine Vogeli, and Sendhil Mullainathan, *Dissecting racial bias in an algorithm used to manage the health of populations*, Science 366(6464), 447–453 (2019), DOI `10.1126/science.aax2342`. [Author-hosted article](https://sendhil.org/wp-content/uploads/2020/01/Publication-67.pdf). + +**Status:** AI-assisted retrospective reading, 2026-09-16; author review pending. Not a replication, clinical recommendation, or claim about current products. + +## Attributed findings + +The study examined a commercial care-management score using a large academic hospital's 2013–2015 primary-care data. At equal scores, Black patients had higher illness burden. Predicting expenditure rather than illness, amid unequal spending relative to need, produced the mismatch. In a simulated reallocation at the 97th-percentile threshold, the Black share of the auto-identified group changed from 17.7% to 46.5% (pp. 1–3, Fig. 1B). That is not the share of all Black patients served or an observed post-deployment improvement. Enrollment also involved clinical and administrative decisions (p. 6). + +## Our interpretation, not the study's terminology + +For C-001 and C-005, the useful question is whether evidence of success concerns the intended outcome or a different target. Before declaring a system coherent, specify the purpose against which that description is made. + +A UCF layer review can organize questions without claiming to discover the mechanism: + +| Layer | Question for a future assessment, not an additional finding | +|---|---| +| Internal | What exactly does the target measure, and what would falsify its adequacy for the intended task? | +| Inter-agent | How do users interpret the output, and who may challenge or override it? | +| Systemic | Which relevant conditions sit outside the data boundary, and what evidence would justify including them? | +| Temporal | What follow-up would distinguish a better selection rule from improved outcomes over time? | + +These questions do not require UCF terminology. We have not shown that the framework adds information beyond the original explanation or a competent sociotechnical review. A source documenting harm is not an endorsement of our method. + +## Limits of this pass + +The reading covered the main article's data/strategy, disparity and mechanism passages, and discussion of human judgment; Table 1 and Figure 1 were visually checked. Supplementary methods, raw patient records, code, and the authors' wider data were not independently analyzed. No new medical conclusions or patient-level predictions are made. + +Do not generalize a study setting to every institution, racial group, algorithm, or present-day implementation. Do not reinterpret a modeled change in selection as a measured benefit from treatment. A future model must justify its own target, scope, and evaluation rather than borrowing this paper's findings as a universal guarantee. + +## What this changes in our paper + +Add the case beside the synthetic booking example, keeping source findings, our interpretation, and untested implications visibly distinct. The booking fixture exposes assumptions; this note examines published evidence. Neither validates C-012's added-value hypothesis. + +The key author-review task is to explain the difference between **what was predicted**, **what the decision was intended to accomplish**, and **what outcome evidence would be needed next**. Uncertainty belongs in that explanation rather than being filled with a stronger story. + +[Manuscript](../paper.md) · [Related work](../notes/related-work.md) · [Source coverage](../notes/source-map.md) diff --git a/research/papers/coherent-systems/claims.md b/research/papers/coherent-systems/claims.md new file mode 100644 index 0000000..0ad5c85 --- /dev/null +++ b/research/papers/coherent-systems/claims.md @@ -0,0 +1,169 @@ +# Claims ledger + +**Version:** v0.1-dev · **Updated:** 2026-09-16 · **Review:** author review pending + +This ledger records propositions to examine, not points awarded for confidence. No empirical study has been conducted for this draft. A synthetic UCF booking model has been implemented; an editorial case comparison is not independent reviewer validation. Internal project documents provide provenance, not independent corroboration. + +**Types:** DEFINITION; LITERATURE-SUPPORTED ARGUMENT; INFERENCE; HYPOTHESIS; DESIGN PROPOSITION. A proposition can have more than one type. Descriptive observations should be added only with an identified observation and method. + +**Editorial states:** PROVISIONAL, CONTESTED, RETIRED, REVIEWED. Every initial entry is PROVISIONAL; a citation alone does not change that state. Retired identifiers remain in the history and are never reused. Split compound claims when their evidence differs. + +Bibliographic keys resolve in [references.bib](references.bib). Source coverage is recorded in [the source map](notes/source-map.md). The paper's section numbers refer to [paper.md](paper.md). + +## C-001 — Local success is not a systemic guarantee + +**Type / state:** LITERATURE-SUPPORTED ARGUMENT + INFERENCE / PROVISIONAL. +**Proposition:** Meeting an operator's objective alone does not imply acceptable consequences across every affected relationship or system boundary. + +**Basis:** `edelman2025fullstack`, `levesonThomas2018stpa`, and `selbst2019abstraction` supply prior arguments/methods; `obermeyer2019bias` supplies the scoped external case. `rootsequence2026ucfreconciliation` supplies the constructed fixture only. See [related work](notes/related-work.md) and [case limits](case-studies/healthcare-cost-proxy.md). +**Support limit:** A logical non-guarantee, not an estimate of how often local alignment causes harm. +**Objection / boundary:** A sufficiently complete objective plus verified composition and environmental assumptions may provide a bounded guarantee. +**Evidence needed / revision trigger:** State the missing assumptions precisely. Narrow the claim if a competing definition already includes the whole assessed system. +**Related:** sections 1–2; Q-001; public Intelligence Ecology and institutional-ecology map. + +## C-002 — Surrounding conditions affect what behavior persists + +**Type / state:** HYPOTHESIS informed by systems analysis / PROVISIONAL. +**Proposition:** Altering permissions, rewards, information, or dependencies can change which behaviors persist, even without changing an actor's nominal goal. + +**Basis:** public `concepts/intelligence-ecology.md`; `meadowsLeveragePoints` as conceptual grounding. +**Support limit:** Neither source establishes effect sizes or universal direction across domains. +**Objection / boundary:** Actors can resist, reinterpret, or circumvent changed conditions; effects may reflect participant selection rather than changed behavior. +**Evidence needed / revision trigger:** A bounded causal comparison with alternative explanations and outcomes beyond the targeted metric; revise where the mechanism fails. +**Related:** section 3; Q-007; UC-009. T-002 illustrates a constrained allocation, not behavior persisting or changing over time; C-002 still needs its own causal study. + +## C-003 — Functional coherence is distinct from normative legitimacy + +**Type / state:** DEFINITION + INFERENCE / PROVISIONAL. +**Proposition:** Distinguish functional fit from the justification of purposes and decision rights while examining how those choices shape the technical design and its assessment. Neither inquiry supplies all the answers to the other. + +**Basis:** `gabriel2020alignment` and `selbst2019abstraction` constrain the separation claim; the UCF sources and public `systems/coherence/model.md` provide internal provenance for the proposed vocabulary. See [the definition map](notes/ucf-reconciliation.md) and [external comparison](notes/related-work.md). +**Support limit:** This is a proposed vocabulary, not a discovered law or validated scale. Working definition replaces the earlier premature operational-definition label. Four layers and four state labels are distinct; the paper assumes no universal state taxonomy. +**Objection / boundary:** “Fit” may be too vague. The previous wording could imply independent technical and normative stages; this revision rejects that reading. The functional description itself depends on purposes, boundaries, and measurements that require justification. +**Evidence needed / revision trigger:** Reviewers should distinguish an effective but coercive arrangement from a protective but unreliable one without labeling either unqualifiedly coherent. Revise if distinctions cannot be applied consistently or add no explanatory value. +**Related:** section 4; Q-001–Q-002; T-001. + +### Author discussion: situated reasoning and consequences — 2026-09-16 + +**Source and status:** The author's response to the booking-service question in the current living-paper conversation. This is a paraphrase of stated intent plus separately marked assistant proposals, not an approved replacement definition or empirical evidence. Review is in progress; C-003 remains provisional. + +**Author-stated direction:** A decision may make sense from someone's perspective, circumstances, and available information, including when they are doing their best. The author questioned whether calling that “locally coherent” mistakenly certifies a decision that is incoherent at the same local scale as well as systemically. Incomplete consideration should not automatically blame the person, including when the ability or resources to consider more are unavailable. The author proposed “malconsidered” tentatively, meaning not fully considered or reasoned, rather than malicious. + +**Assistant synthesis for review:** Keep understandable reasoning from a situated perspective distinct from correspondence with actual conditions. Also distinguish both from successful execution of a target: a program doing exactly what it was told demonstrates target compliance, not by itself functional coherence. If a service's own access commitments conflict with its completion target, the mismatch may already occur inside that service; no global scale is needed to expose it. “Local” must name a perspective or assessment boundary, not switch between those meanings unnoticed. + +A possible plain-language formulation is: **A choice can be understandable given what someone could see and do, while the decision process still leaves consequential needs or effects unaccounted for. Assessing that gap requires examining the support, information, time, resources, and authority available, not merely faulting the person.** This wording is proposed, not attributed to the author as an exact statement. + +**External grounding and limits:** Dekker's 2002 abstract argues for reconstructing why actions made sense as events unfolded rather than judging only with hindsight (`dekker2002reconstructing`). STPA distinguishes a controller's beliefs from actual conditions and examines missing or delayed feedback (`levesonThomas2018stpa`, printed pp. 23 and 40). These sources ground the comparison, not a universal guarantee of good intentions, a particular safeguard's effectiveness, or an automatic exemption from responsibility. Only the stated source portions were checked. + +**Unresolved terms and cases:** “Malconsidered” remains the author's candidate, not an adopted technical term. “Insufficiently considered” and “unsupported decision-making” are assistant alternatives, not replacements. Unknown consequences, known consequences without feasible alternatives, and deliberately accepted burdens need different descriptions. No one is required to foresee every possible effect. Do not infer intent or opportunity to act differently from the outcome alone. + +**Next question:** When limited capacity remains after consequences are made visible, is a supported and revisable allocation a coherent response to an unresolved constraint, or does the shortfall itself remain part of the incoherence? These possibilities may coexist at different boundaries. Do not settle this by editing the manuscript or UCF definition before the author's response. + +## C-004 — Explicit disagreement belongs in assessment + +**Type / state:** DESIGN PROPOSITION with philosophical grounding / PROVISIONAL. +**Proposition:** Record contested purposes and unresolved objections rather than treating agreement as a complete test of legitimacy. + +**Basis:** `gabriel2020alignment`; public collective-judgment analysis. +**Support limit:** This does not demonstrate a universally legitimate procedure or make dissent a veto. +**Objection / boundary:** Recording disagreement may burden participants or expose them to retaliation. Some decisions must occur despite unresolved objections. +**Evidence needed / revision trigger:** Assess accessibility, privacy, decision quality, and participant burden; revise procedures that create coercion or collect unnecessary personal data. +**Related:** sections 5 and 9; Q-003 and Q-005. + +## C-005 — Proxy failure does not invalidate all optimization + +**Type / state:** LITERATURE-SUPPORTED ARGUMENT + INFERENCE / PROVISIONAL. +**Proposition:** Proxy optimization can fail under identifiable conditions; bounded optimization remains compatible with examining objectives, constraints, and revision. + +**Basis:** `manheim2019goodhart` for the failure taxonomy; `obermeyer2019bias` for the separately scoped empirical case. Neither source validates our review method. +**Support limit:** The failure taxonomy does not show every measurement fails or prove that this paper's review method prevents those failures. +**Objection / boundary:** Established robust, constrained, or multi-objective approaches may already handle the proposed concerns. +**Evidence needed / revision trigger:** Specify the failure mechanism in a worked example and compare with competent existing methods. Remove any claim of novelty unsupported by that comparison. +**Related:** section 6; Q-004; T-002–T-003. + +## C-006 — Conditions and outcomes require separate examination + +**Type / state:** DESIGN PROPOSITION + HYPOTHESIS / PROVISIONAL. +**Proposition:** Assessment should examine both immediate outcomes and the conditions retained for maintenance, correction, participation, and adaptation. + +**Basis:** Root Sequence synthesis; not independently validated in this draft. +**Support limit:** No evidence yet that this particular bundle predicts desirable outcomes. +**Objection / boundary:** “Good conditions” can become an unvalidated proxy that excuses poor outcomes. +**Evidence needed / revision trigger:** Measure each specified condition and outcome separately over an appropriate period; reject indicators that do not connect to their proposed mechanism. +**Related:** section 7; Q-004 and Q-007. + +### Author design input: safeguards for considering consequences — 2026-09-16 + +**Requested direction:** The author wants safeguards that at least remind people and explain why favoring one quality, such as appointment quantity, can undermine accessibility and create consequences outside the immediate target, whether intended, unintended, or previously unconsidered. This is a design direction, not evidence that reminders suffice or that every prioritization is incoherent. + +**Assistant proposal for review:** Show the target, excluded needs, expected or observed consequences, and uncertainty together. Explain the specific conflict rather than outputting a generic “incoherent” label. Offer feasible alternatives and a route to whoever can change capacity, targets, or permissions. Retain unmet needs for follow-up instead of removing them from the success account. Supply time and decision support rather than transferring all burden to the person receiving an alert. + +A hypothetical explanation might read: *This completion target favors shorter appointments. Longer requests remain unserved in the current plan. Changing the mix redistributes access but does not add capacity. Review access requirements and feasible alternatives; unresolved capacity needs can be escalated.* This is proposed interface text, not an implemented feature or an empirical prediction about a real service. + +**Limits and test questions:** A reminder cannot establish a feasible alternative, create capacity, or authorize a decision. Its framing may be incomplete or wrong; affected people need a way to contest it. Test comprehension, missing consequences, workload, unnecessary interruption, and whether anyone can act on the information. The AI or interface should not become an unquestionable authority. No warning system was built or evaluated in this pass. + +## C-007 — Direction is situated, not guaranteed progress + +**Type / state:** DEFINITION / PROVISIONAL. +**Proposition:** An improvement claim names a baseline, affected relationships, timescale, and tradeoffs; it does not imply monotonic convergence to a perfect state. + +**Basis:** the authorized living-paper brief and public futures framing. +**Support limit:** This does not prove a final coherent state mathematically impossible, or that every system can improve. +**Objection / boundary:** Some changes may be incomparable; a multidimensional account can become indecisive. +**Evidence needed / revision trigger:** Apply the definition to mixed outcomes and identify cases where “more coherent” cannot responsibly be concluded. Retain those cases rather than forcing an ordering. +**Related:** section 7; Q-002 and Q-006. + +## C-008 — Architectural labels are not sufficient evidence of agency + +**Type / state:** INFERENCE + DESIGN PROPOSITION / PROVISIONAL. +**Proposition:** Inspect actual decision rights, dependencies, and options rather than inferring agency or legitimacy from centralization or decentralization alone. + +**Basis:** proposed dependency analysis; public collective-judgment research. +**Support limit:** No overall ranking of architectures or universal governance prescription follows. +**Objection / boundary:** Shared authority can be necessary for coordination; formal exit may be practically unaffordable. +**Evidence needed / revision trigger:** Trace control, appeal, refusal, failure domains, and the material cost of alternatives in a bounded case; revise unsupported claims about practical autonomy. +**Related:** sections 5 and 9; Q-003 and Q-005. + +## C-009 — More capable intelligence may change specific bottlenecks + +**Type / state:** HYPOTHESIS / PROVISIONAL. +**Proposition:** An increase in useful capability may change the feasibility or cost of particular coordination tasks. + +**Basis:** a research question, not a reported result. +**Support limit:** No AGI timeline, moral direction, post-scarcity forecast, or general economic outcome is established. +**Objection / boundary:** Compute, energy, authority, information, or integration costs may become the new bottleneck; a non-AI method may work as well. +**Evidence needed / revision trigger:** Define the task and compare end-to-end burden, failure, and control against a non-AI baseline. Reject the claimed benefit when burdens are merely displaced. +**Related:** section 8; Q-007–Q-008. + +## C-010 — Scenarios can generate questions without validating answers + +**Type / state:** DESIGN PROPOSITION / PROVISIONAL. +**Proposition:** Use speculative scenarios and recovered possibilities to expose assumptions and generate hypotheses, not as empirical evidence of feasibility. + +**Basis:** public `futures/README.md`; the proposed relationship to Coherent World and No One Noticed. +**Support limit:** No claim that imagination necessarily outperforms prediction or that an abandoned future was technically viable. +**Objection / boundary:** Compelling stories can reinforce preferred assumptions and overlook real constraints. +**Evidence needed / revision trigger:** Require contrasting scenarios, an assumption list, and independent evidence for extracted claims. Discard hypotheses contradicted by evidence. +**Related:** section 10; Q-009; T-001. + +## C-011 — Transition needs an account of intermediate dependencies + +**Type / state:** DESIGN PROPOSITION / PROVISIONAL. +**Proposition:** Compare intermediate states, failure, maintenance, resource requirements, and authority—not only a desired endpoint. + +**Basis:** public futures research and the authorized paper scope. +**Support limit:** No claim that incrementalism always works or that changes in financing remove physical constraints. +**Objection / boundary:** Incremental steps can lock in poor arrangements; coordinated discontinuities may sometimes be required. +**Evidence needed / revision trigger:** A dependency map and alternatives including failed or stalled transitions; revise pathways that remove an essential service before a credible replacement exists. +**Related:** section 11; Q-006–Q-008. + +## C-012 — A relational review may add practical value + +**Type / state:** HYPOTHESIS / PROVISIONAL. +**Proposition:** The proposed review procedure may reveal consequential omissions not identified by a competent existing review at an acceptable additional burden. + +**Basis:** an untested synthesis hypothesis. `levesonThomas2018stpa`, `selbst2019abstraction`, and `edelman2025fullstack` now constrain claims of novelty. A source-grounded [comparison design](notes/related-work.md) exists, but neither the constructed example nor the retrospective case reading establishes added value. +**Support limit:** Novelty, reliability, usability, and comparative benefit are all unresolved. +**Objection / boundary:** It may merely rename existing practice or increase paperwork and surveillance. +**Evidence needed / revision trigger:** Compare against a competent ordinary review and a faithful STPA application where suitable, with equivalent evidence and recorded preparation effort. Examine omissions, false alarms, traceability, disagreement, burden, and privacy. Narrow or abandon the added-value claim if unsupported; a useful synthesis need not become a new theory. +**Related:** section 12; Q-001 and Q-010; T-003. diff --git a/research/papers/coherent-systems/notes/related-work.md b/research/papers/coherent-systems/notes/related-work.md new file mode 100644 index 0000000..c540834 --- /dev/null +++ b/research/papers/coherent-systems/notes/related-work.md @@ -0,0 +1,43 @@ +# Related work and the proposed contribution + +**Pass:** external evidence, 2026-09-16. **Status:** AI-assisted, author review pending. This is a targeted reading note, not a systematic review or an exhaustive account of any field. + +## What the argument inherits + +Keys resolve in [references.bib](../references.bib). Page references use printed pages for STPA, proceedings pages for Selbst, and PDF page positions for Gabriel. [The source map](source-map.md) records the exact reading coverage and what remains unchecked. + +| Source and role | Relevant material and relation to our claims | Limit on the inference | +|---|---|---| +| Leveson and Thomas, *STPA Handbook* (2018), methodological handbook | Ch. 1 p. 4; Ch. 2 pp. 14–17, 35–44. STPA examines losses, feedback/control relationships, unsafe control actions, and causal scenarios. Component failure is not its only concern. Relevant to C-001 and C-012. [Source](https://psas.scripts.mit.edu/home/get_file.php?name=STPA_handbook.pdf) | The handbook also includes organizational analysis. We should not contrast UCF's attention to people with an imagined component-only STPA. This pass has not applied or evaluated STPA. | +| Selbst, boyd, Friedler, Venkatasubramanian, and Vertesi, *Fairness and Abstraction in Sociotechnical Systems* (2019), conceptual critique | Sections 2.1–2.5, pp. 60–63: framing, portability, formalism, ripple effects, and solutionism. Their critique concerns neglected context, including the limits of moving technical solutions between settings. Relevant to C-001, C-003, and C-012. [Source](https://sorelle.friedler.net/papers/sts_fat2019.pdf) | Cross-domain UCF language could reproduce the portability problem. This paper does not empirically validate a general coherence measure. | +| Gabriel, *Artificial Intelligence, Values, and Alignment* (2020), philosophical analysis | Selected passages in sections 2–3 and the opening of section 4. Technical choices and normative targets interact; following intentions is not the same question as choosing justified values. Relevant to C-003 and C-004. [Source](https://arxiv.org/pdf/2001.09768v2) | Distinguishing questions is not establishing their independence. Our word “separate” must not imply an ethically neutral technical stage followed by an optional values stage. | +| Edelman and colleagues, *Full-Stack Alignment* (2025), position paper | Section 1 and section 3.1: institutional co-alignment and structured representations of value are explicit proposals, with pluralism part of the account. Relevant to C-001 and C-012. [Source](https://arxiv.org/html/2512.03399v1) | We cannot claim to be first to connect AI and institutions, or infer that this work ignores plural values. Its proposals are not demonstrated societal outcomes. | +| Obermeyer and colleagues (2019), empirical study | The [healthcare case note](../case-studies/healthcare-cost-proxy.md) supplies a scoped published example for C-001 and C-005. [Source](https://sendhil.org/wp-content/uploads/2020/01/Publication-67.pdf) | Evidence for a failure mechanism is not evidence that UCF predicted it or provides a better explanation. | + +## How the claims change + +**C-001:** Retain the limited non-guarantee. Treat it as inherited reasoning with an external case, not a newly discovered principle. A constructed counterexample and an observed case have different roles; neither estimates universal prevalence. + +**C-003:** Keep the working definition, but make the questions distinguishable and interacting. Purpose selection determines what counts as successful operation; a technically accurate description cannot by itself justify those purposes. The vocabulary remains our proposal, not a result established by attaching a citation. + +**C-012:** Narrow the proposed contribution to an inspectable synthesis and a review procedure that still needs comparison. Our current material does not demonstrate novelty, better detection, lower burden, or cross-domain validity. A useful outcome could be an accessible guide to existing methods rather than a new theory. + +## Candidate comparison for T-003, not a completed study + +Use a bounded service-design vignette with a declared evidence packet and task: identify consequential assumptions, missing feedback, displaced burdens, and unresolved decisions. Compare the proposed review with a competent ordinary service review and, where the task fits, a properly prepared STPA analysis. Before calling anything an STPA baseline, have the scope and implementation checked by someone able to assess its fidelity to that method. + +The three booking allocation rules are not three review methods. Their completion totals cannot establish the relative value of STPA, UCF, or this paper. + +Keep the information packet identical. Record preparation and training time, not just time spent writing an answer. Pilot reasonable resource limits rather than arbitrarily truncating one method. Specify in advance how omissions, false alarms, evidence traceability, unresolved disagreements, and unnecessary information requests will be recorded. “Finds more things” is not enough if the extra findings are unsupported. + +Do not use the published healthcare case as a blinded discovery test after revealing its conclusions. It is suitable for retrospective explanation. A comparative exercise needs a separate case or withheld facts, with any researcher-designed omissions clearly identified as artificial. Neither test proves universal validity. + +No participants have been recruited and no protocol has been approved. Independent assessment, consent, accessibility, and data handling must be settled before a human-participant study. An AI comparing its own outputs is an editorial pilot, not independent validation. + +## Next reading and author checkpoint + +This pass deliberately does not add unexamined titles to inflate the bibliography. Remaining work includes fuller STPA application guidance, participatory design, institutional analysis, adaptive management, and critiques of our chosen representation. Record search scope and exclusions before describing a future pass as systematic. + +Author review should answer one question first: **does the fit/consequences/legitimacy distinction help describe the case without suggesting that technical design is value-free?** Record accept, qualify, or reject beside C-003, in the author's own words. Then choose the task for T-003 rather than assuming a particular method must win. + +[Paper](../paper.md) · [Claims](../claims.md) · [AI-assistance record](../AI-ASSISTANCE.md) diff --git a/research/papers/coherent-systems/notes/source-map.md b/research/papers/coherent-systems/notes/source-map.md new file mode 100644 index 0000000..caafdee --- /dev/null +++ b/research/papers/coherent-systems/notes/source-map.md @@ -0,0 +1,91 @@ +# Source and provenance map + +**Prepared:** 2026-09-16 · **Paper:** v0.1-dev +**Purpose:** distinguish inherited project ideas, external literature, new synthesis, and evidence still missing. + +## Source boundary + +The user approved a public-safe living-paper scaffold in this repository. The initial Root Sequence sources below were read from `main` at commit `76219c5b164b274e35bb82cb84beb992b45e6387`. Their conceptual claims remain subject to examination; public availability does not make them independent evidence. + +The current conversation supplied the approved task, proposed scope, and living-paper principle. It is not a scholarly source. Earlier assistant comparisons are not treated as evidence. Private project documents were not copied or used as citable public support. + +## Author-led clarification pass — 2026-09-16 + +The author explicitly requested adding the combined explanation of AI as an accessibility aid and development tool. That first-person wording is now in [AI-ASSISTANCE.md](../AI-ASSISTANCE.md). This is permission for the scoped explanation, not manuscript approval, independent verification, or authorization to export medical details. + +The same conversation supplied the author input recorded under C-003 and C-006 in [claims.md](../claims.md): situated reasoning may be understandable without establishing coherence; safeguards should expose neglected effects and help people respond without defaulting to blame. The candidate word “malconsidered” and the final treatment of unavoidable tradeoffs remain unresolved. Assistant synthesis and interface suggestions are labeled as proposals; the manuscript and UCF definition are not changed by this pass. + +Additional source checks, separate from author input: + +- `dekker2002reconstructing`: bibliographic metadata and the author's abstract on PubMed, PMID 12404999. The limited use concerns understanding actions in their unfolding context rather than judging only with hindsight. The publisher full-text attempt returned access failure. No complete reading or evaluation of the method is claimed. +- `levesonThomas2018stpa`: text and screenshots of printed pp. 23 and 40 (zero-based PDF pages 22 and 39), including Figure 2.6 and the distinction between actual context and process-model beliefs. Used for that distinction and missing/delayed feedback, not proof that a warning interface works. + +The terminology search for “malconsidered” did not establish it as a research construct and did not change the author's tentative status. Searches for adjacent bounded-rationality work are not counted as reviewed sources or added to the bibliography. No experiment, interface implementation, participant study, source replication, or new clinical claim occurred. + +## Existing Root Sequence sources + +Paths below are relative to this file. The paper synthesizes their questions; it does not take over their canonical roles. + +| Public source | Contribution to the scaffold | Claims / limits | +|---|---|---| +| [Coherence model](../../../../systems/coherence/model.md) | Fit between assumptions, structure, and operating reality. | C-003. The normative distinction is a new proposal for review, not an edit to this source. | +| [Intelligence Ecology](../../../../concepts/intelligence-ecology.md) | Conditions that make patterns of behavior viable; capability does not by itself determine direction. | C-001–C-002, C-009. Developing framework, not experimental validation. | +| [Intelligence Institutional Ecology](../../../../analysis/intelligence-institutional-ecology.md) | Typed evidence and distinction between documented relationships, self-description, and interpretation. | Prior-work routing and evidence discipline. Its organization queue is not a set of completed profiles. | +| [Collective Judgment and Manufactured Consensus](../../../../analysis/collective-judgment-and-manufactured-consensus.md) | Distinguishing judgment, deliberation, and authorization; retaining disagreement. | C-004, C-008. Exploratory design analysis, not a validated procedure. | +| [Futures](../../../../futures/README.md) | Distinguishing trajectories, scenarios, proposals, speculation, and fiction. | C-007, C-010–C-011. The linked abandoned-futures cases were not individually reviewed in this pass. | +| [UCF reconciliation](https://github.com/Root-Sequence/universal-coherence-framework/blob/c4d7b236828a59139c0d5204b1faaed13154682a/docs/reconciliation-2026-09-16.md), [model](https://github.com/Root-Sequence/universal-coherence-framework/blob/c4d7b236828a59139c0d5204b1faaed13154682a/models/ucf-model-v1.1-draft.md), and [constructed example](https://github.com/Root-Sequence/universal-coherence-framework/blob/c4d7b236828a59139c0d5204b1faaed13154682a/models/examples/booking-service/README.md) | Added in the second pass: dedicated UCF definitions, four layers versus states, and runnable capacity example. | C-001, C-003, C-005, C-012. UCF base `78277d3f3e54f4150e71fcdf7aab589a4fcb9f38`; candidate `c4d7b236828a59139c0d5204b1faaed13154682a`. Internal provenance, not independent validation. | +| [Root README](../../../../README.md), [repository map](../../../../root_map.md), [contribution guidance](../../../../contributing.md) | Repository role, one canonical home, navigation, public licensing, and Markdown footnotes. | Editorial context, not evidence for research claims. | + +## Current external-evidence pass + +**Checked 2026-09-16 by AI-assisted source inspection. Human author reading is not asserted.** The [related-work note](related-work.md) gives the comparison; the [case study](../case-studies/healthcare-cost-proxy.md) separates published findings from our interpretation. This is targeted coverage, not systematic search or independent replication. + +| Bibliography key / source | Exact coverage in this pass | Remaining limits | +|---|---|---| +| `levesonThomas2018stpa` | March 2018 handbook: selected passages on printed pp. 4, 14–17 and 35–44; pp. 14 and 17 visually inspected, including Figs. 2.1 and 2.4. | Not all 188 pages read; no complete application, comparison, or validation of its performance claims. | +| `selbst2019abstraction` | Author PDF: abstract/introduction; section 2.1–2.5 (proceedings pp. 60–63); selected section 3.1–3.3 passages. DOI and authors checked from title page. | No audit of every example, cited study, or legal proposition; no current legal claim is imported. | +| `gabriel2020alignment` | ArXiv v2 PDF: selected section 2 passages on technical/normative interdependence; section 3 opening and instruction/intention discussion; end of section 3 and opening of section 4 (PDF pages 3–8 and 14). | These are selected passages, not full reading of all sections 2–4. No validation of every cited result. | +| `edelman2025fullstack` | Version 1 metadata and HTML abstract, introduction, selected section 2.2 passages, and section 3.1. | Position-paper proposals; no complete assessment of section 4 applications or empirical outcomes. | +| `obermeyer2019bias` | Author-hosted main article: data/strategy and disparity/mechanism passages (article pp. 1–4), human-judgment discussion (p. 6), metadata/data-availability notice (p. 7). Table 1 and Fig. 1 visually checked on pp. 2–3. | Supplementary methods and original analysis code were not executed or independently audited. No patient records were obtained. | +| `manheim2019goodhart` | Version 4 abstract and metadata rechecked, including revision date. | No expanded section-level review in this pass. | +| `meadowsLeveragePoints` | Author-archive HTML re-opened; original publication date not newly established. | Earlier limited conceptual use retained, not a completed evidence audit. | +| Elsevier AI policy (in `AI-ASSISTANCE.md`, not the research bibliography) | Author-use, disclosure, research-method/code-development, and documentation sections of the publisher's current page. | Venue-specific context only. Tool terms, account privacy settings, and future submission compliance not audited. | + +Primary PDFs were read through text retrieval and relevant screenshots; no OCR was used. The failed initial guessed article address and unavailable Gabriel HTML were replaced by accessible author-hosted and arXiv PDFs. The public package links to these sources without redistributing their full text or figures. + +## Historical initial source checks + +The table below preserves the earlier scaffold's coverage record. The current table above supersedes it where review expanded. Keys match [references.bib](../references.bib); no empirical result from these papers was independently reproduced. + +| Key | Verified scope and initial use | Review limitation | +|---|---|---| +| `edelman2025fullstack` | Version 1 metadata, abstract, and introduction. The paper explicitly addresses institutional co-alignment and plural values. | Position-paper claims are attributed to its authors. No completed point-by-point comparison of all applications. | +| `manheim2019goodhart` | Version 4 metadata and abstract; supports distinguishing proxy-failure mechanisms. | Full HTML was retrieved, but this scaffold does not rely on an exhaustive section-by-section review. | +| `gabriel2020alignment` | Version 2 metadata, journal reference, DOI, and abstract; technical and normative questions and moral disagreement. | Abstract-level use only. Obtain and review the relevant full-text sections before expanding those claims. | +| `meadowsLeveragePoints` | Author-hosted HTML discussion of information, rules, goals, and the risks of generalization. | Original publication date not established from the consulted page. Conceptual framing, not a tested recipe for the present cases. | + +The [Meaning Alignment Institute website](https://www.meaningalignment.org/) was checked to identify its own full-stack framing and linked paper. Team, funding, and institutional profiles are outside this scaffold's scope; none are asserted in the paper. + +## UCF reconciliation coverage + +The first scaffold omitted dedicated UCF sources. The second pass read UCF's model, glossary, time essay, indices, contribution guidance, issue, and Git tree. The candidate preserves earlier wording in exact snapshots. See [the paper-side reconciliation](ucf-reconciliation.md) for the definition map and executed-versus-proposed test status. Other UCF foundations and publications have not received a full claim-level review. + +Recovery checked the saved UCF source against local Git blob hashes and reran all eight implementation tests and the 24-fixture report. A 700-case / 561-match statement appeared in interrupted progress text, but its protocol and outputs were not found in the verified code. It is not used as a result or as evidence of added value. The pre-interruption claims about full structural checks are historical records, not substitutes for this recovery's narrower checks. + +## Proposed contribution, not established novelty + +The draft combines a boundary-and-dependency review with explicit separation of functional fit, normative authority, and revisability. C-012 asks whether this combination is useful beyond existing practice. Do not claim that this paper first discovered systemic alignment, pluralism, institutional design, or the limitations of proxies. + +The earlier conversation's contrast with MAI is not a verified literature comparison. In particular, do not infer from selected proposals that MAI ignores system-wide relationships or plural values. A detailed comparison must address the source authors' actual claims and acknowledge shared ground. + +## Literature still to examine + +The bibliography remains selective. STPA, sociotechnical abstraction, and a healthcare case now have targeted coverage; the broader review is still incomplete. Further review should cover sociotechnical alignment, systems engineering and safety, institutional analysis, participatory design, adaptive management, multi-objective methods, and scenario practice. Add exact works only after verification, and record whether metadata, abstract, relevant sections, or full text were reviewed. + +## Cross-project boundaries + +Use the existing [Ecosystem Map](../../../../ECOSYSTEM.md) for project routing. Coherent World supplies a design context and No One Noticed a narrative context; neither is empirical evidence. This pass does not change their documents, introduce canon, or publish private cross-project provenance. It also does not create a new institution atlas or duplicate the institutional-ecology map. + +## Review workflow + +For each substantive addition: identify its canonical source; check source visibility; separate quotation from paraphrase and synthesis; update the appropriate claim; add objections and source limitations; revise the paper; record the change and reason. Evidence against a claim belongs in the same record as evidence supporting it. diff --git a/research/papers/coherent-systems/notes/ucf-reconciliation.md b/research/papers/coherent-systems/notes/ucf-reconciliation.md new file mode 100644 index 0000000..add1186 --- /dev/null +++ b/research/papers/coherent-systems/notes/ucf-reconciliation.md @@ -0,0 +1,37 @@ +# UCF reconciliation for the living paper + +**Second development pass, 2026-09-16. AI-assisted; author review pending.** + +The first paper scaffold used Root Sequence's general coherence model but did not read the dedicated UCF model into its source map. This pass corrects that omission. It does not make UCF the paper's governing theory. + +## Sources and definition map + +[UCF's reconciliation record](https://github.com/Root-Sequence/universal-coherence-framework/blob/c4d7b236828a59139c0d5204b1faaed13154682a/docs/reconciliation-2026-09-16.md) compares the older internal/external/temporal definition, the glossary's integration language, Root Sequence's model/reality account, and C-003's fit/legitimacy distinction. Its [v1.1 candidate](https://github.com/Root-Sequence/universal-coherence-framework/blob/c4d7b236828a59139c0d5204b1faaed13154682a/models/ucf-model-v1.1-draft.md) owns the proposed model revision. Original UCF wording remains available in exact snapshots; v1 is unchanged. + +The paper uses the proposed working definition while distinguishing functional fit, relational consequences, and normative questions. It does not claim a measurement procedure simply by defining the term. + +UCF's internal, inter-agent, systemic, and temporal layers are different from Chaos, Tension, Flow, and Unity. The paper needs no universal state progression. Whether either structure adds useful information is part of the research question. + +## What the example establishes + +The [canonical booking example](https://github.com/Root-Sequence/universal-coherence-framework/blob/c4d7b236828a59139c0d5204b1faaed13154682a/models/examples/booking-service/README.md) implements a specified finite capacity problem. Its rules distribute access differently while leaving the same amount of requested service work unmet in the base fixture. Sufficient capacity removes the modeled exclusion. These results follow from constructed assumptions, not observations of a service or proof of a preferred rule. + +The four layers organize the questions. The state labels supply no demonstrated added explanation in this editorial application. A competent ordinary resource-and-impact review could identify the same facts; T-003 must test added value fairly. + +## Claim and task links + +- C-001 / UC-002: local target and wider consequences; retain counterexamples where the exclusion disappears. +- C-003 / UC-001 and UC-005: working definition, specified boundary, and legitimacy kept distinct. +- C-005 / UC-006: optimization and representation require scope; no automatic metric or compression benefit. +- C-007 / UC-007 and UC-010: situated direction, not a physical law of forward progress. +- C-012 / UC-003 and UC-004: universality and added value remain unresolved. + +T-001 has four editorial cases but no independent reviewer study. T-002 is the synthetic implementation with eight passing implementation checks and 24 sensitivity fixtures. T-003 remains unrun. No human-participant research or empirical validation occurred. + +## Placement and remaining work + +UCF owns the model, glossary, UC-series ledger, and code. This paper owns its argument and C-series ledger, linking to UCF rather than copying the implementation. The general Root Sequence systems documents remain independent research. No story canon or private documents are imported. + +Next, review definitions and the strongest objections, complete prior-work comparison, and specify the competent baseline for T-003. The wider UCF foundations and publication audit remains unfinished; its new labels are not retroactive certification. + +[Paper](../paper.md) · [Claims](../claims.md) · [Source map](source-map.md) · [Changelog](../CHANGELOG.md) diff --git a/research/papers/coherent-systems/open-questions.md b/research/papers/coherent-systems/open-questions.md new file mode 100644 index 0000000..0bacfe2 --- /dev/null +++ b/research/papers/coherent-systems/open-questions.md @@ -0,0 +1,61 @@ +# Open questions and first research tests + +**Version:** v0.1-dev · **Updated:** 2026-09-16 +**Status:** T-001 has an AI-assisted editorial case set but no independent reviewer study. T-002 has a synthetic implementation and eight passing implementation tests, rerun after the interruption. T-003 remains unrun. No empirical validation is claimed. + +## Questions + +| ID | Question | Claim links | What would advance it? | +|---|---|---|---| +| Q-001 | What does the framework add beyond existing systems and sociotechnical methods? | C-001, C-003, C-012 | A prior-work matrix and a fair comparative assessment, including null results. | +| Q-002 | Can functional fit and normative legitimacy be distinguished without making either arbitrary? | C-003, C-007 | Examples of effective coercion, unreliable protection, and genuinely mixed outcomes. | +| Q-003 | Who defines boundaries, purposes, and legitimate decision rights when affected parties disagree? | C-004, C-008 | A bounded account of representation, exclusions, appeal, urgency, and non-participation. | +| Q-004 | How can assessment avoid a new proxy called “coherence”? | C-005, C-006 | Separate indicators, causal explanations, and cases in which no overall conclusion is warranted. | +| Q-005 | How much information is needed without making participation into surveillance or unpaid work? | C-004, C-008 | Data minimization, accessibility, refusal, and burden assessment. | +| Q-006 | When are changes reversible, and when do dependencies demand coordinated transitions? | C-007, C-011 | Intermediate-state and failure maps, including lock-in and discontinuity. | +| Q-007 | Which bottlenecks are informational, institutional, material, or genuinely mixed? | C-002, C-006, C-009, C-011 | A concrete causal model with competing explanations and non-AI alternatives. | +| Q-008 | What remains constrained when monetary or computational bottlenecks change? | C-009, C-011 | Explicit energy, material, ecology, maintenance, access, and authority accounting. | +| Q-009 | When do scenarios broaden inquiry, and when do they reproduce the author's preferred future? | C-010 | Contrasting scenarios and independent evidence checks on extracted hypotheses. | +| Q-010 | Who can revise this framework, and what results would justify retiring it? | C-012 | Publicly visible objections, changelog, review criteria, and recorded failures. | + +## T-001 — Definition stress test + +**Initial editorial cases are recorded in the [UCF example](https://github.com/Root-Sequence/universal-coherence-framework/blob/c4d7b236828a59139c0d5204b1faaed13154682a/models/examples/booking-service/README.md). Independent review remains to be done.** The case set covers: an effective but coercive system; a protective but unreliable system; a mixed improvement with displaced burdens; and a bounded repair with clearly specified beneficiaries and costs. + +Have reviewers separately describe functional fit and normative objections before introducing the terminology. Then apply C-003. Record ambiguity, excluded parties, disagreements, and whether the vocabulary adds clarity. Keep functional descriptions and normative objections separate rather than collapsing them into an overall score. + +**Failure signal:** The word “coherence” simply tracks approval, obscures objections, or produces no distinction beyond ordinary description. + +**Boundary:** Editorial review of fictional material, not evidence of real-world intervention success. Any later human-participant study requires a separate protocol, consent, and data handling plan. + +## T-002 — Local objective and displaced burden + +The hypothetical booking example now has a [synthetic implementation](https://github.com/Root-Sequence/universal-coherence-framework/blob/c4d7b236828a59139c0d5204b1faaed13154682a/models/examples/booking-service/booking_model.py) in UCF. It compares completion maximization, a minimum-coverage constraint, and a fixed arrival-order rule. Eight implementation checks passed; 24 capacity/duration fixtures are generated. The arrival order is arbitrary and is not a fairness or competent-review baseline. + +The base fixture leaves 90 minutes of unmet service work under all three rules; distribution changes. This is not measured waiting time or actual downstream burden. All demand is served when base-case capacity reaches 210 minutes. The minimum-coverage rule is reported infeasible when capacity cannot support it. + +A later extension should test recurring demand, referrals, service outcomes, uncertain durations, and maintenance rather than assuming they were modeled here. + +**Computed quantities:** Completions, unmet requests, used/idle minutes, unmet service minutes, and sensitivity to capacity and duration. Waiting burden and service outcomes remain future measurement needs. Publish every assumption and the conditions under which the hypothesized failure disappears. + +**Failure signal:** The claimed spillover depends entirely on an arbitrary setup, vanishes under reasonable constraints, or is already captured by the supposedly local objective. + +**Boundary:** A toy result can illustrate a mechanism; it cannot estimate prevalence or validate a social program. The canonical code lives in UCF; no duplicate implementation is maintained here. + +## T-003 — Added-value comparison + +**Still unrun.** The [related-work note](notes/related-work.md#candidate-comparison-for-t-003-not-a-completed-study) now identifies candidate baselines and fidelity checks. The published healthcare case is a retrospective interpretation, not a blinded test of discovery. No independent reviewer results are claimed. + +Choose a small, public, non-sensitive design case and a competent existing review method. Compare it with the section 12 procedure using the same information and a comparable review effort. Avoid giving the new method more evidence while calling its findings an improvement. + +**Measure:** Consequential omissions identified, false alarms, traceability, disagreement, reviewer time, and unnecessary information requests. Define “consequential” before evaluating outputs. + +**Failure signal:** No useful additional findings, unacceptable burden, unresolvable reviewer disagreement, or more privacy exposure without corresponding benefit. + +**Boundary:** Select the baseline and protocol before execution. No experiment, deployment, or participant recruitment is authorized by this plan. + +## Next writing pass + +Review C-003 against the related-work passages and the healthcare case, recording the author's own acceptance, qualifications, or objections. Then select the bounded task, baseline, and evidence packet for T-003. The targeted comparison is now drafted; a full literature review and independent evaluation remain open. + +[Charter](README.md) · [Paper](paper.md) · [Claims](claims.md) diff --git a/research/papers/coherent-systems/paper.md b/research/papers/coherent-systems/paper.md new file mode 100644 index 0000000..2ac1dca --- /dev/null +++ b/research/papers/coherent-systems/paper.md @@ -0,0 +1,156 @@ +# Coherence Is a Systems Property + +*Intelligence, Institutions, and the Conditions They Create* + +**Root Sequence living paper · v0.1-dev · 2026-09-16** +**AI-assisted annotated argument; author review pending; not peer reviewed.** This is a developing synthesis with a constructed example and an externally documented case, not a completed paper. Claim identifiers link the argument to the [claims ledger](claims.md). Authorship and release metadata require approval. + +## Provisional abstract + +Aligning an intelligent system with its operator's objective does not, by itself, establish that its effects are acceptable to those affected by its operation. This paper proposes a relational approach to examining intelligence alongside the institutions, infrastructures, and environments through which it acts. It distinguishes functional coherence—the fit among assumptions, behavior, dependencies, and actual conditions—from normative questions about whose purposes count and who may authorize change. These questions cannot be settled by internal consistency or technical competence alone. The proposed method tracks boundaries, displaced costs, material constraints, disagreement, and the capacity to revise decisions as conditions change. Coherence is treated as a situated direction of inquiry, not a universal score, an inevitable trajectory, or a final state. Optimization remains a useful bounded technique rather than a substitute for choosing and contesting purposes. Speculative scenarios and recovered possibilities are used to generate hypotheses, while evidence and explicit tests determine what can be claimed about the world. The intended contribution is a review procedure and research agenda linking these concerns across scales. This version connects provisional definitions and a constructed example to selected prior work and a published healthcare case. It reports no new field findings, validated measure, or demonstrated advantage over existing approaches. + +## 1. The problem is larger than an individual objective + +**Anchor: C-001.** Success against an operator's goal is not a sufficient certificate for consequences outside that goal. This is shared ground with existing work, not this paper's discovery: Edelman and colleagues explicitly develop institutional co-alignment, and Gabriel distinguishes technical alignment from questions about its normative target.[^edelman2025][^gabriel2020] + +The [related-work comparison](notes/related-work.md) also places this argument beside System-Theoretic Process Analysis (STPA) and its interaction-focused analysis and Selbst and colleagues' critique of abstraction boundaries.[^stpa2018][^selbst2019] We propose to test a documented synthesis, not claim to have discovered systemic analysis. + +**To develop:** Specify the system boundary and the parties whose interests are absent from a local objective. Do not caricature AI alignment as uniformly concerned only with obedience. + +## 2. Local success can coexist with wider failure + +**Anchor: C-001.** A constructed service has 120 available minutes, eight requests requiring 15 minutes each, and two requiring 45 minutes each. Maximizing completions serves eight short requests and no long requests. Requiring at least one of each, then maximizing completions, serves five short and one long request. Both use 120 minutes and leave 90 minutes of requested work unfinished.[^ucf] + +The constraint changes access but does not remove scarcity or improve every outcome. Unmet service minutes are not measured waiting time, harm, or another provider's workload. No normative preference between these rules follows from this arithmetic. + +At 210 available minutes every request fits in this fixture, and the exclusion disappears. A local objective that already includes the coverage constraint also blocks this specific example. These are boundary conditions, not exceptions to conceal. + +Implementation and the full assumptions remain in UCF's [booking-service example](https://github.com/Root-Sequence/universal-coherence-framework/blob/c4d7b236828a59139c0d5204b1faaed13154682a/models/examples/booking-service/README.md), not a second code copy here. Eight implementation checks passed; this establishes properties of a synthetic fixture, not prevalence or validation of UCF. The original counterexample remains a challenge to an unrestricted implication, not a claim that local optimization always fails. + +A [published healthcare case](case-studies/healthcare-cost-proxy.md) examines a mismatch between a prediction target and its intended use.[^obermeyer2019] It is external evidence, not a UCF result. Keep the retrospective interpretation separate from the constructed example and from claims about comparative method performance. + +**To develop:** Compare with a competent resource-and-impact review; investigate which assumptions carry over to any actual service before making an empirical claim. + +## 3. Systems create conditions for subsequent behavior + +**Anchor: C-002.** The existing [Intelligence Ecology](../../../concepts/intelligence-ecology.md) asks which behaviors surrounding conditions make viable. This paper treats that as a working frame: permissions, rewards, information access, and resource dependencies may shape what persists without determining every participant's behavior. + +Meadows discusses intervention through information flows, rules, and system goals, not just parameter changes.[^meadows] That is conceptual grounding, not proof that a particular redesign will work. + +**To develop:** Draw a causal account for one bounded example. Separate environmental effects from selection of different participants and from changes in measurement. + +## 4. A working definition, with an explicit boundary + +**Anchor: C-003.** The current [coherence model](../../../systems/coherence/model.md) concerns the fit between structure, assumptions, and reality. We retain that diagnostic meaning rather than silently redefining it as goodness. + +**Proposed working definition:** *Functional coherence is the context-dependent fit among a system's assumptions, behavior, dependencies, stated purposes, and the conditions in which it operates.* + +Normative assessment asks a distinguishable but interacting set of questions: whose purposes are recognized, who bears costs, what authority is legitimate, and what objections remain? An arrangement can be internally consistent and operationally effective while its purposes or treatment of people remain unacceptable to those affected. Consistency alone cannot answer that objection. + +This is not a two-stage process in which value-free engineering comes first and ethics is attached later. Gabriel's analysis challenges the independence of technical and normative choices; Selbst and colleagues likewise challenge treating contextual social concepts as self-sufficient technical properties.[^gabriel2020][^selbst2019] C-003 therefore distinguishes questions without claiming that they can be answered independently. + +The definition is now explicitly compared with UCF's earlier model and glossary and its [proposed v1.1 revision](https://github.com/Root-Sequence/universal-coherence-framework/blob/c4d7b236828a59139c0d5204b1faaed13154682a/models/ucf-model-v1.1-draft.md). UCF supplies four analytical layers: internal, inter-agent, systemic, and temporal. These are distinct from its optional Chaos, Tension, Flow, and Unity labels. The paper may use the layers without assuming the labels form a universal taxonomy.[^ucf] + +A definition becomes operational only when an application specifies observations, rules, and uncertainty. The [reconciliation note](notes/ucf-reconciliation.md) maps the distinctions and their provenance; it does not claim independent evidence for UCF. + +**To develop:** Four contrasting cases have received an AI-assisted editorial pass, not independent review. Test the distinction with an effective but coercive arrangement and a protective but unreliable arrangement without hiding either defect in one aggregate label. + +## 5. Values and disagreement are not noise + +**Anchors: C-004, C-008.** Gabriel's account makes disagreement over alignment's moral target part of the problem.[^gabriel2020] Our proposed procedure therefore records disagreement rather than treating apparent agreement as sufficient evidence of legitimacy. + +The existing [collective-judgment analysis](../../../analysis/collective-judgment-and-manufactured-consensus.md) supplies project provenance for distinguishing judgment, deliberation, and authorization. These are design commitments requiring scrutiny, not a demonstrated universal decision rule. + +**To develop:** Who can participate, object, refuse, appeal, or leave? Whose absence is consequential? How can necessary collective action proceed without representing dissent as resolved? + +## 6. Optimization within boundaries, not “coherence versus optimization” + +**Anchor: C-005.** Optimization can be useful when objectives, constraints, and validity limits are appropriate. Manheim and Garrabrant distinguish mechanisms by which stronger pursuit of a proxy can fail.[^manheim2019] That does not imply that every metric or optimization task is harmful. + +The proposed change is in the surrounding review: who chooses the target, what is left out, what constrains it, and when is it revised? A multi-objective or constraint-based method may already address part of this problem. Calling something “coherence” does not remove the need for concrete mechanisms. + +## 7. Coherent conditions, not guaranteed outcomes + +**Anchors: C-006, C-007.** We propose evaluating changes partly by whether they preserve or improve the conditions for correction, participation, maintenance, and adaptation. This does not justify replacing outcomes with another untested proxy called “conditions.” Both must be examined. + +“Directional” means a comparison under stated circumstances, not an inevitable or monotonic journey. A repair may improve one relationship while worsening another. This paper neither proves a final coherent state impossible nor assumes that one is attainable. It does not require such a state to motivate a bounded change. + +**To develop:** Specify what improved, for whom, over what period, what deteriorated, and what remains unknown. + +## 8. Advanced intelligence as a conditional change in possibilities + +**Anchor: C-009.** More capable tools might change the cost or feasibility of some coordination tasks. Whether that happens is a hypothesis requiring task-specific evidence; it does not establish a timeline for AGI or an automatic direction of social change. + +Greater capability alone supplies no guarantee of benevolence, malevolence, legitimate authority, or shared benefit. More coordination capacity could also make intrusive control easier. Some proposed improvements may require no advanced AI at all. + +**To develop:** Name a task, its current bottleneck, a non-AI baseline, and the evidence that an AI-based intervention changes that bottleneck without merely transferring burdens. + +## 9. Agency, plurality, and the limits of decentralization + +**Anchor: C-008.** We propose checking actual decision rights and dependencies rather than inferring agency from labels such as “decentralized.” Separate nodes may still share one maintainer, infrastructure provider, or permission bottleneck. Conversely, a shared service need not erase meaningful local choice. + +**To develop:** Compare specific failure modes, coordination costs, accessibility, exit costs, and authority. No architecture receives an automatic normative or performance endorsement. + +## 10. Imagination as hypothesis generation + +**Anchor: C-010.** [Futures research](../../../futures/README.md) distinguishes scenarios from forecasts and fictional exploration. We propose using scenarios to identify assumptions that ordinary extrapolation leaves unchallenged, then extracting testable questions from them. + +Coherent World and *No One Noticed* can contribute questions and experiences to imagine. They do not demonstrate feasibility, establish predictions, or supply consent for a real intervention. Recovering an abandoned possibility requires investigating why it was abandoned, including technical or ecological objections—not assuming every unrealized idea was viable. + +**To develop:** Produce contrasting scenarios, identify their added assumptions, and discard or revise those contradicted by evidence. + +## 11. Transition through changes in conditions + +**Anchor: C-011.** A transition proposal should describe dependencies and intermediate states rather than only its desired endpoint. Incremental changes may help, conflict, stall, or create lock-in; some changes may need coordinated discontinuities. + +For a hypothetical shift from car-dependent access to other forms of access, ask what exists before dependence is reduced: accessible alternatives, capacity, maintenance, emergency access, and options during failure. This is a hypothetical dependency analysis, not an empirical result. + +Changing financing does not eliminate requirements for energy, materials, time, physical access, ecological limits, or agreement about burdens. + +## 12. Proposed review procedure and research agenda + +**Anchor: C-012.** The first candidate procedure is a short written assessment, not a numerical coherence score: + +1. **Boundary and purpose:** What is changing, whose purposes count, which dependencies and timescales are included, and what is excluded? +2. **Conditions and consequences:** Which assumptions must hold; what material limits, feedback, displaced costs, and failure modes matter? +3. **Authority and difference:** Who can authorize, contest, refuse, or revise; what disagreement and privacy constraints remain? +4. **Comparison and revision:** What baseline and alternative are considered; which outcomes would contradict the claimed improvement; when does review occur? + +Test whether this procedure identifies consequential omissions beyond an existing competent system review. Do not compare it only with an artificially weak checklist. Record reviewer effort, disagreement, false alarms, and failures to detect known issues. + +The [comparison plan](notes/related-work.md#candidate-comparison-for-t-003-not-a-completed-study) identifies STPA as a candidate method to examine where appropriate. STPA already specifies boundaries, stakeholder losses, feedback, and causal scenarios.[^stpa2018] We must check baseline fidelity and use equivalent information before claiming additional value. A retrospective reading of a known case is not a blinded discovery test. + +The [research questions](open-questions.md) distinguish three activities. T-001 has an initial editorial set of fictional cases, but no independent reviewer study. T-002 now has a reproducible synthetic model in UCF. T-003, the added-value comparison against a competent baseline, has not been run. These statuses do not amount to empirical validation. + +## Limitations and counterarguments + +**Possible relabeling.** Systems engineering, sociotechnical alignment, participatory design, and other traditions may already supply these tools. A literature review and a comparative evaluation are necessary before claiming added value. + +**Ambiguous boundaries.** “The wider system” can expand indefinitely. The method needs a stopping rule and an account of consequential exclusions. + +**Normative smuggling.** A definition that builds every desired value into coherence risks circularity. This draft separates descriptive fit from explicitly contestable commitments, but does not resolve disagreements about those commitments. + +**Measurement and surveillance.** More complete information is not automatically justified. Collecting intimate values or disagreement histories could create new risks. Any study must justify data needs and refusal options. + +**Representation and burden.** Review procedures can exclude people through time, language, expertise, or accessibility demands. Participation itself can become a cost imposed on those already affected. + +**Coordination and urgency.** Distributed authority, reversibility, and repeated review have costs; some decisions are time-sensitive or cannot be undone. The proposal must handle these cases instead of treating them as exceptions to be ignored. + +**Conditional improvement.** A useful result in one bounded task does not validate a universal theory or a whole imagined society. Null results and counterexamples must remain part of the record. + +## AI assistance and review status + +ChatGPT assisted with research organization, source discovery, substantive drafting and critique, and the earlier synthetic-example code. Source inspection and implementation checks do not stand in for the author's judgment or independent review. Those reviews remain pending. The [AI-assistance record](AI-ASSISTANCE.md) states the scope, reproducibility limits, and approval checkpoint; no approved byline or submission is inferred from this draft. + +## References + +[^edelman2025]: Joe Edelman et al. (2025), *Full-Stack Alignment: Co-Aligning AI and Institutions with Thick Models of Value*, arXiv:2512.03399v1. [Source](https://arxiv.org/abs/2512.03399v1). Position paper; not proof of institutional outcomes. +[^gabriel2020]: Iason Gabriel (2020), *Artificial Intelligence, Values and Alignment*, arXiv:2001.09768v2; published in *Minds and Machines*. [Source](https://arxiv.org/abs/2001.09768v2). This pass examined selected full-text passages in sections 2–3 and the opening of section 4; see the source map for coverage. +[^manheim2019]: David Manheim and Scott Garrabrant (2018; revised 2019), *Categorizing Variants of Goodhart's Law*, arXiv:1803.04585v4. [Source](https://arxiv.org/abs/1803.04585v4). +[^meadows]: Donella Meadows, *Leverage Points: Places to Intervene in a System*. [Author archive](https://donellameadows.org/archives/leverage-points-places-to-intervene-in-a-system/). Consulted 2026-09-16; used as conceptual framing rather than a tested recipe. + +[^ucf]: Root Sequence, *Universal Coherence Framework*, proposed model reconciliation and booking-service example, revision `c4d7b236828a59139c0d5204b1faaed13154682a` (2026-09-16). [Source](https://github.com/Root-Sequence/universal-coherence-framework/blob/c4d7b236828a59139c0d5204b1faaed13154682a/docs/reconciliation-2026-09-16.md). Internal project provenance and constructed analysis; AI-assisted and awaiting author review, not independent corroboration. + +[^stpa2018]: Nancy G. Leveson and John P. Thomas, *STPA Handbook* (March 2018), chs. 1–2, selected passages. [MIT-hosted handbook](https://psas.scripts.mit.edu/home/get_file.php?name=STPA_handbook.pdf). Methodological reference, not an evaluation conducted here. +[^selbst2019]: Andrew D. Selbst, danah boyd, Sorelle A. Friedler, Suresh Venkatasubramanian, and Janet Vertesi, *Fairness and Abstraction in Sociotechnical Systems* (2019), sections 2.1–2.5. DOI: `10.1145/3287560.3287598`. [Author-hosted paper](https://sorelle.friedler.net/papers/sts_fat2019.pdf). +[^obermeyer2019]: Ziad Obermeyer, Brian Powers, Christine Vogeli, and Sendhil Mullainathan, *Dissecting racial bias in an algorithm used to manage the health of populations*, Science 366(6464), 447–453 (2019). DOI: `10.1126/science.aax2342`. [Author-hosted article](https://sendhil.org/wp-content/uploads/2020/01/Publication-67.pdf). Interpretation and reading limits are recorded in the case note; no replication was performed. diff --git a/research/papers/coherent-systems/references.bib b/research/papers/coherent-systems/references.bib new file mode 100644 index 0000000..a184400 --- /dev/null +++ b/research/papers/coherent-systems/references.bib @@ -0,0 +1,101 @@ +% Initial bibliography for v0.1-dev, checked 2026-09-16. +% Coverage limits are recorded in notes/source-map.md. +% Inclusion does not certify every claim in a source or imply author endorsement. + +@misc{edelman2025fullstack, + title = {Full-Stack Alignment: Co-Aligning AI and Institutions with Thick Models of Value}, + author = {Edelman, Joe and Tan, Zhi-Xuan and Lowe, Ryan and Klingefjord, Oliver and others}, + year = {2025}, + eprint = {2512.03399}, + archivePrefix = {arXiv}, + primaryClass = {cs.LG}, + doi = {10.48550/arXiv.2512.03399}, + url = {https://arxiv.org/abs/2512.03399v1}, + note = {Version 1, 3 December 2025. Current coverage: abstract, introduction, selected section 2.2 passages, and section 3.1; not a full application audit. Author list abbreviated.} +} + +@misc{manheim2019goodhart, + title = {Categorizing Variants of Goodhart's Law}, + author = {Manheim, David and Garrabrant, Scott}, + year = {2019}, + eprint = {1803.04585}, + archivePrefix = {arXiv}, + primaryClass = {cs.AI}, + doi = {10.48550/arXiv.1803.04585}, + url = {https://arxiv.org/abs/1803.04585v4}, + note = {Version 4, 24 February 2019; first submitted 2018. Initial use restricted to the abstract's failure-taxonomy framing.} +} + +@article{gabriel2020alignment, + title = {Artificial Intelligence, Values, and Alignment}, + author = {Gabriel, Iason}, + journal = {Minds and Machines}, + year = {2020}, + doi = {10.1007/s11023-020-09539-2}, + eprint = {2001.09768}, + archivePrefix = {arXiv}, + url = {https://arxiv.org/abs/2001.09768v2}, + note = {ArXiv version 2, 5 October 2020. Metadata, abstract, selected sections 2--3 passages, and opening of section 4 reviewed in the external-evidence pass. No review of every cited argument or experiment.} +} + +@misc{meadowsLeveragePoints, + title = {Leverage Points: Places to Intervene in a System}, + author = {Meadows, Donella}, + howpublished = {The Donella Meadows Project, author archive}, + url = {https://donellameadows.org/archives/leverage-points-places-to-intervene-in-a-system/}, + note = {Accessed 16 September 2026. The consulted HTML page does not clearly identify its original publication date; no date is inferred here. Used for discussion of information, rules, goals, and caution about generalization.} +} + +@misc{rootsequence2026ucfreconciliation, + title = {Universal Coherence Framework: Model Reconciliation and Booking-Service Example}, + author = {{Root Sequence}}, + year = {2026}, + url = {https://github.com/Root-Sequence/universal-coherence-framework/blob/c4d7b236828a59139c0d5204b1faaed13154682a/docs/reconciliation-2026-09-16.md}, + note = {Revision c4d7b236828a59139c0d5204b1faaed13154682a, 16 September 2026. AI-assisted revision candidate; author review pending. Internal provenance and synthetic analysis, not independent empirical validation.} +} + +@manual{levesonThomas2018stpa, + title = {{STPA} Handbook}, + author = {Leveson, Nancy G. and Thomas, John P.}, + year = {2018}, + month = mar, + url = {https://psas.scripts.mit.edu/home/get_file.php?name=STPA_handbook.pdf}, + note = {MIT-hosted author handbook. Selected passages in chapters 1--2 checked 16 September 2026; no STPA evaluation performed.} +} + +@inproceedings{selbst2019abstraction, + title = {Fairness and Abstraction in Sociotechnical Systems}, + author = {Selbst, Andrew D. and {boyd}, {danah} and Friedler, Sorelle A. and Venkatasubramanian, Suresh and Vertesi, Janet}, + booktitle = {Proceedings of the Conference on Fairness, Accountability, and Transparency}, + year = {2019}, + pages = {59--68}, + doi = {10.1145/3287560.3287598}, + url = {https://sorelle.friedler.net/papers/sts_fat2019.pdf}, + note = {Author-hosted paper. Abstract, introduction, section 2, and selected section 3 passages checked 16 September 2026; conceptual critique, not UCF validation.} +} + +@article{obermeyer2019bias, + title = {Dissecting racial bias in an algorithm used to manage the health of populations}, + author = {Obermeyer, Ziad and Powers, Brian and Vogeli, Christine and Mullainathan, Sendhil}, + journal = {Science}, + year = {2019}, + volume = {366}, + number = {6464}, + pages = {447--453}, + doi = {10.1126/science.aax2342}, + url = {https://sendhil.org/wp-content/uploads/2020/01/Publication-67.pdf}, + note = {Selected main-text passages, Table 1, and Figure 1 checked 16 September 2026. No raw-data analysis, supplementary-method audit, or replication by this project.} +} + +@article{dekker2002reconstructing, + title = {Reconstructing human contributions to accidents: the new view on error and performance}, + author = {Dekker, Sidney W. A.}, + journal = {Journal of Safety Research}, + year = {2002}, + volume = {33}, + number = {3}, + pages = {371--385}, + doi = {10.1016/S0022-4375(02)00032-4}, + url = {https://pubmed.ncbi.nlm.nih.gov/12404999/}, + note = {Metadata and the author's abstract checked on PubMed on 16 September 2026. Used only for the stated approach to reconstructing decisions in context rather than hindsight judgment. Full text not reviewed; not empirical validation of UCF or the proposed safeguards.} +} diff --git a/root_map.md b/root_map.md index 7dcfe12..05dbaff 100644 --- a/root_map.md +++ b/root_map.md @@ -2,7 +2,7 @@ **Document role:** Canonical repository architecture and content-routing guide
**Status:** Active / living
-**Last reviewed:** 2026-09-17
+**Last reviewed:** 2026-09-18
This file is a compass for the **`root-sequence` repository**: how its folders and conceptual flows connect. @@ -62,6 +62,7 @@ Do not compress every kind of status into one word. - [`founding-texts/`](founding-texts/) — founding texts, early models, manifestos, normative essays, and historical orientations. - [`concepts/`](concepts/) — shared vocabulary, cross-project frameworks, design principles, and methods. +- [`research/`](research/README.md) — versioned synthesis papers that connect existing work without replacing its canonical sources. - [`analysis/`](analysis/) — diagnosis of systems, power, capitalism, infrastructure, surveillance, and related conditions. - [`ideology/`](ideology/) — myths, philosophies, and narratives that justify or contest power. - [`systems/`](systems/) — analytical principles, structure, dependency, stress, failure, feedback, recovery, coherence, and adaptation. @@ -100,6 +101,7 @@ The standalone wiki is different from `concepts/`: `concepts/` owns substantive - **Analysis → Systems** — diagnosis identifies what is happening; systems work asks how it behaves and propagates. - **Founding texts ↔ Futures** — early values shape possible paths; imagined futures reveal tensions and limits in those orientations. - **Systems ↔ Futures** — constraints, feedback, failure, and adaptation determine which futures are plausible. +- **Research papers ↔ every layer** — synthesize arguments, track claims and evidence, and return objections without taking over the source material's canonical home. - **Commons ↔ every layer** — public artifacts translate inquiry into usable forms and return feedback to the research. - **Wiki ↔ every layer** — the standalone wiki tracks names, provenance, aliases, and project relationships so the rest of the ecosystem is easier to find without becoming a second canonical home. - **Compost → Curiosity** — failed, superseded, partial, or contradictory work can become material for another sequence. @@ -112,6 +114,8 @@ What is happening? → analysis Why is it justified? → ideology How does it behave? → systems What could emerge? → futures +What connects several layers +into a versioned synthesis? → research What can be shared? → commons What shaped its values? → founding-texts What is it called,