Manifold Grace
Research · Information Science · Public Knowledge Systems

Information Science
& Network Visibility

Investigating how interconnected public identities, original work and evidence networks become more discoverable, credible and durable — and how their downstream effects can be assessed probabilistically.

Research framework · First published 18 October 2026 · Methodology and proposed measures, not a claim of current site performance.

Research question

What cascades — and what lasts?

How can a coherent public knowledge identity be structured to maximise relevant discovery and useful secondary connections while preserving authorship, provenance, interpretability and independent verification?

Original work → Persistent record → Connected profiles → Independent references → Qualified discovery → Long-term use

Distribution is not the same as validation. A self-authored profile establishes what its author says; credible external references, verifiable professional evidence and independently evaluated publications contribute different forms of corroboration.

Research progress ·

Portable evidence and accountable retrieval

Developed a reference design linking authorised source records, independent review, minimal approved attestations and permission-aware AI retrieval. The proposal separates source authenticity from claim accuracy and preserves uncertainty, contrary evidence, version history and corrections.

Mapped a ten-step signal path from secure ingestion to a qualified, cited answer and later correction. A bounded pilot could use existing storage and read-only APIs, with tool adapters and a shared schema; a central lake of private records is not required.

Concept and outreach completed; no operational registry or independent validation claimed. Read the concise proposal record · Dated progress log.

Interdisciplinary application · 3 November 2026

Evidence accuracy in criminal justice: an illustrative application

The proposed authorised-records and accountable-retrieval framework may also help legal professionals evaluate case information without substituting AI for evidence or professional judgment. For example, an authorised reviewer could associate an assertion with its original record, event date, source institution, later amendments and any conflicting material; distinguish corroborated, disputed and unverified claims; and flag missing records for proportionate follow-up.

Employment, education and public-body histories are examples of fragmented records that can matter in different kinds of legal inquiries. A historical “as-of” check could distinguish the policy in force at an event date from a later revision. Any application to criminal defence or investigation would require lawful access, source-preserving audit trails, confidentiality safeguards and qualified human review. This is a proposed use case, not an operational forensic tool or an assessment of any particular law firm.

Related: AI quality assurance · Research context · Dated research contribution · Feature proposals and correspondence.

Interdisciplinary foundations

Four intersecting fields

Network Science

Nodes, edges, centrality, clustering, bridging ties and the diffusion of information between people, documents and organisations.

Webometrics

Measurement of online information resources, hyperlinks, web visibility, referring domains and discoverability.

Scientometrics

Measurement of scholarly production, citations, authorship, collaboration and research influence; bibliometrics supplies publication-level techniques.

Digital Identity Management

Consistent identifiers, entity resolution, profile linkage, provenance, identity claims and privacy-aware management of public information.

Information retrievalKnowledge graphsBibliometricsStochastic processesInformation diffusionSearch visibility

Measurement · observed vs unavailable

Evidence dashboard

This dashboard retrieves selected public records at viewing time and lets you inspect a Google Search Console export locally. Search Console metrics require site-owner access; this public webpage does not request passwords, API tokens or OAuth permissions, and does not upload CSV files.

Public-source observations

GitHub repository

Public stars—
Public forks—
Latest repository push—

Not yet retrieved

Stars and forks are platform-specific engagement indicators, not measures of identity authority.

OpenAlex publication lookup

Pilot DOI: 10.5281/zenodo.17439725 · indexed records may differ from the repository's own metadata.

OpenAlex indexed citations—
Publication year (index)—

Work title not retrieved

Not yet retrieved

Search Console · local-only assessment

In Search Console, open Performance → Search results, choose an explicit date range, export the Pages or Queries table as CSV, then select the file here. Exported rows can be limited or aggregated. Results reflect only the rows provided, not necessarily all website traffic.

Clicks in imported rows—
Impressions in imported rows—
Calculated CTR—
Impression-weighted position—

Awaiting local Search Console CSV export

No CSV data is sent to Manifold Grace. Values exist only in the current page session and disappear on reload. Do not enter private account credentials into a public page.

Independent reference register · demonstration

Enter an external reference for inspection. Entries appear only in the current browser view, are not published, and are not automatically validated as endorsements.

No user-added references in this session

ReferenceClassification

Data quality: All figures are attributed by source. A dash (—) means not retrieved, not provided or unavailable; it does not mean zero. The public metrics depend on third-party API access and may fail under rate limits. No automatic Search Console access is configured.

Operational framework

Metrics that can be assessed

Each metric needs a defined population, observation window, source and denominator. The following are assessment options, not currently measured values. Platform-specific metrics are not always directly comparable.

DimensionMetric / operational definitionHow to measureInterpretive caution
Identity integrityCross-profile consistency rate = matching verified identity fields ÷ assessed fieldsQuarterly profile audit: name, description, identifiers, official linksAgreement among self-controlled profiles is not independent verification
Entity connectivityVerified profile-link coverage = profiles linked to canonical identity ÷ relevant profiles auditedWebsite crawl; ORCID and other linked accountsOnly publish links intentionally made public
Search discoveryOrganic impressions, relevant queries, indexed canonical pages, organic CTRGoogle Search Console, with page/query/date filtersImpressions do not equal readers or endorsement
Audience qualityQualified engagement rate = relevant actions ÷ eligible visitsConsent-appropriate analytics; enquiries, document reads, contact clicksDefine a meaningful action before calculating
Web citation networkUnique independent referring domains; contextual editorial referencesReferral logs, backlink tools and manual source reviewPurchased, duplicated or automated links should be excluded
Scholarly outputWorks with persistent identifiers; publication and version completenessDOI registry, repository metadata, ORCID works recordDOI registration does not imply peer review
Scholarly receptionIndependent citations per work, time-normalised where appropriateCrossref/OpenAlex/discipline-specific citation records; manual deduplicationCoverage differs by discipline and source
Collaboration networkUnique substantive co-authors, organisations and documented joint outputsAuthorship records and project evidenceMembership or casual interaction is not necessarily collaboration
Graph structureDegree and betweenness centrality; connected components and bridgesConstruct an attributed graph of public profiles, works and referencesHigh centrality may reflect volume rather than quality
Knowledge reuseIndependent adaptations, educational reuse, downloads and references per assetRepository statistics and attributable reuse recordsDownloads cannot reliably identify unique readers
DurabilityLink survival rate = accessible canonical resources ÷ previously published resourcesRepeat HTTP checks and archive/identifier auditReachability alone does not prove ongoing relevance
Economic conversionRelevant enquiries, opportunities and attributable outcomes per publication or channelVoluntary enquiry logs; source-coded contact routesAvoid treating correlation as proof of causation
Equity & accessibilityAccessible-page coverage, keyboard usability, alt-text and readability checksAutomated plus manual WCAG-oriented reviewAutomated scores cannot establish full accessibility
Privacy & controlPublic-data minimisation, correction latency, documented consent and removal routesPrivacy audit and change logsMore public personal information is not inherently better

What should be published with any reported number?

Reporting period; collection date; dataset scope; numerator and denominator; measurement tool; inclusion and exclusion rules; known missing data; and whether the number is observed, estimated, modelled or independently verified.

Probability and network effects

Model stochastic cascades

A simplified branching model estimates how many subsequent relevant discoveries may follow from an initial exposure. It is for comparing assumptions, not predicting platform algorithms, rankings or personal success.

Expected next-generation discoveries = current relevant discoveries × average onward shares × probability of a relevant follow-on discovery
Reproduction factor0.24
Additional discoveries expected31
Total across generations131

Illustrative expected values only. Formula: R = onward shares × relevant discovery probability; expected total = initial discoveries × (1 + R + … + R^g). Assumes constant parameters, independent propagation and no audience overlap or saturation. These assumptions seldom hold perfectly in real networks.

Choose actions by expected value — not visibility alone

Indicative decision score = probability of relevant reach × value of outcome × persistence × onward-reference potential ÷ time and cost

This heuristic has no universal calibrated weights. Compare scenarios against observed outcomes, estimate uncertainty ranges and revisit after each reporting period. Where evidence allows, use controlled comparisons or attribution analysis rather than assuming every new enquiry came from a single post.

Publishing architecture

High-value durable content forms

Evidence-based case studies

Document problem, work performed, artefacts, verification status and outcome. Useful for employers and collaborators.

Research papers and data

Persistent identifiers, methods, versioning, sources and appropriate rights statements support citation and reuse.

Reference explainers

Glossaries, tutorials, diagrams and policy briefs connect specialist work to broader audiences.

Author and project records

Stable biographies, structured metadata, project pages and reciprocal official links help readers navigate the identity network.

Publication cascade: primary work → canonical landing page → professional summary → visual explainer → scholarly or professional discussion → verifiable external references. Each derivative should point to the primary source and disclose substantive changes.

Research protocol

Build a baseline, then test

  1. Inventory: record publicly intended profiles, URLs, identifiers and core authored works.
  2. Baseline: measure indexing, referrals, citations, accessible pages and meaningful enquiries for a defined period.
  3. Graph: map identity → works → organisations → independent references; mark evidence and uncertainty for each edge.
  4. Intervene: improve one category at a time — e.g. metadata, original publication, collaboration or distribution.
  5. Compare: track changes over consistent windows and account for seasonality, platform changes and confounding events.
  6. Report: publish methodology and a dated table, including blanks where data are not yet available.

Current measurement status: no analytics accounts or scholarly APIs are connected to this page. No impressions, citations, conversion rates, centrality scores or probabilities are presented as observed results. The simulator is local and does not collect user data.

Primary methodological references

Reading and measurement sources

These are reference organisations and documentation links, not endorsements of the Manifold Grace methodology. This page proposes an interdisciplinary assessment design rather than claiming that the full model has been empirically validated.

← Return to Research · Research Collaborations · Professional

Cross-representational evidence research

A complementary study examines machine-readable claims extracted from charts, diagrams and documents, with versioned sources, semantic alignment and independently checked synthesis. Read the proposed research framework.

Additive research note, 4 November 2026. Conceptual and testable; not a validated ASI capability.