{"schema_version":1,"research_id":"eoa_inverse_innovation_exp05_external_evaluation_20260803","source_assessment_id":"invariant_mode_decomposition_design__literature_literary_theory:P5:v0","cell_id":"invariant_mode_decomposition_design__literature_literary_theory","search_queries":["site:unesco.org oral traditions archive repeated performances safeguarding consent community participation ethics","site:paradisec.org collection protocol community access oral tradition recordings","Mukurtu Traditional Knowledge Labels cultural heritage archives permissions community protocol official","computational analysis oral tradition narrative variation repeated performances motif sequence study","site:paradisec.org about archive endangered languages music performance recordings access protocols","site:elararchive.org depositors access protocols endangered languages archive recordings official","oral tradition repeated performances same narrative variation performance primary research computational","folktale computational phylogenetic narrative variation motifs prior art research","ELAR archive access protocols official depositors community endangered languages","PARADISEC protocols access conditions official collection recordings","Endangered Languages Archive access protocols deposit official","oral history archives sampling collection gaps diversity archivist expressed need","dynamic mode decomposition text narratives language documents topic evolution","dynamic mode decomposition natural language text analysis arxiv","spectral analysis narrative sequences literature computational narratology PCA","dynamic mode decomposition small sample noise stability official documentation","BLS archivists curators museum workers median pay 2025 archivist official","BLS anthropologists archeologists median pay 2025 official","research assistant hourly wage university digital humanities 2026 official","oral history project budget recording transcription cost guidance","UNESCO Twelve Ethical Principles safeguarding intangible cultural heritage endorsed 2015 date","Library of Congress Ancestral Voices Passamaquoddy rights access publication date","Archives and oral history Oral History Society publication date"],"sources":[{"source_id":"S1","title":"Ethics and Intangible Cultural Heritage","publisher":"UNESCO Intangible Cultural Heritage","url":"https://ich.unesco.org/en/ethics-and-ich-00866","source_class":"OFFICIAL_GUIDANCE","publication_date":"2015-12-04","accessed_at":"2026-08-03","claims_supported":["Communities, groups, and individuals should have the primary safeguarding role.","Heritage research and documentation should use transparent collaboration and free, prior, sustained, informed consent.","Community moral and material interests in research and documentation must be protected."]},{"source_id":"S2","title":"Rights and Access: Ancestral Voices","publisher":"Library of Congress, American Folklife Center","url":"https://memory.loc.gov/collections/ancestral-voices/about-this-collection/rights-and-access/","source_class":"GOVERNMENT_OR_REGULATOR","publication_date":"n.d.","accessed_at":"2026-08-03","claims_supported":["The Library of Congress and Passamaquoddy community implemented a collaborative archival project with culturally sensitive curatorial protocols.","The Passamaquoddy Tribe retains cultural authority over its heritage and defines responsible-use guidance.","Access for education and research does not eliminate the user's responsibility to obtain necessary further permissions."]},{"source_id":"S3","title":"Archives and Oral History","publisher":"Oral History Society","url":"https://ohs.org.uk/archives-and-oral-history/","source_class":"OFFICIAL_GUIDANCE","publication_date":"2022","accessed_at":"2026-08-03","claims_supported":["Archives sponsor interviews to fill collection gaps and undertake storage, digitization, cataloguing, transcription, finding-aid, and access work.","Archives can fund, support, and partner with communities, academics, and practitioners.","The professional community identifies underused or little-known sound and oral-history collections as a concern."]},{"source_id":"S4","title":"Oral History Good Practice Guidance","publisher":"The National Lottery Heritage Fund","url":"https://www.heritagefund.org.uk/funding/good-practice-guidance/oral-history","source_class":"OFFICIAL_GUIDANCE","publication_date":"2024-01-29; updated 2025-11-24","accessed_at":"2026-08-03","claims_supported":["A heritage funder identifies filling historical gaps, documenting inherited stories, and recording changing and enduring community culture as legitimate needs.","Projects should consult communities and archives to establish demand, avoid duplication, and enhance collections.","Recording agreements, data-protection bases, copyright, licensing, redaction, embargo, safeguarding, deposit agreements, and professional documentation are workflow requirements.","Time-coded content summaries and links to recordings are recognized archival practices.","Project costs depend on staff, training, research, travel, translation, transcription, documentation, archiving, equipment, outputs, and evaluation."]},{"source_id":"S5","title":"Story Alteration in Oral History Retellings: Methods of Comparative Work","publisher":"John Benjamins Publishing Company, Narrative Inquiry","url":"https://benjamins.com/catalog/ni.24018.mly","source_class":"PRIMARY_RESEARCH","publication_date":"2025-01-23","accessed_at":"2026-08-03","claims_supported":["Digital archives can contain multiple tellings of the same episode by the same narrator under different circumstances.","Narrative alteration reflects content, surrounding life-story structure, interactional setting, and archival situatedness.","Comparability is an analytical achievement requiring context, not an automatic property of repeated recordings."]},{"source_id":"S6","title":"A Method to Represent Variations of Narratives Using a Motif-Network Model","publisher":"Japanese Cognitive Science Society","url":"https://www.jstage.jst.go.jp/article/jcss/12/4/12_4_338/_article/-char/en","source_class":"PRIMARY_RESEARCH","publication_date":"2005","accessed_at":"2026-08-03","claims_supported":["Prior research represents variants of a story through motifs and a network of narrative progression.","The model was demonstrated on Cinderella variants and supports reconstruction of narrative structure.","Explicit computational modeling of coupled narrative variation therefore predates this proposal, although not for rights-governed sampling allocation."]},{"source_id":"S7","title":"Representations Matter: Embedding Modes of Large Language Models Using Dynamic Mode Decomposition","publisher":"arXiv","url":"https://arxiv.org/abs/2309.01245","source_class":"PRIMARY_RESEARCH","publication_date":"2023-09-03","accessed_at":"2026-08-03","claims_supported":["Dynamic mode decomposition has been applied to text embeddings by treating sentence position as time and embedding coordinates as state.","The study uses spectra, retained modes, low-rank approximation, and reconstruction-related contrasts on text.","The demonstrated application concerns sentence embeddings and LLM hallucination, not successive living performances or archival sampling."]},{"source_id":"S8","title":"Designing a Project","publisher":"U.S. National Park Service","url":"https://home.nps.gov/articles/000/oral-history-resources-designing-a-project.htm","source_class":"GOVERNMENT_OR_REGULATOR","publication_date":"n.d.","accessed_at":"2026-08-03","claims_supported":["Oral-history projects require explicit goals, resource assessment, role assignment, repository planning, and consultation with prospective narrators.","A complete interview can require approximately 30–60 labor hours, and transcription may cost about $125–$150 per recorded hour.","Existing comparable archives should be checked before collecting new interviews."]}],"problem_evidence":{"support":"WEAK","rationale":"External evidence establishes that archives deliberately fill gaps, that repeated tellings vary with discourse and context, and that preserving changing and enduring community traditions matters. It does not demonstrate the proposal's narrower failure: an archive with many recordings systematically missing a stable coupled transformation mode because it sampled by performer, occasion, motif, or visible difference. No prevalence, retrospective failure, or downstream-harm study for that mechanism was found.","source_ids":["S3","S4","S5"]},"stakeholder_evidence":{"support":"MODERATE","rationale":"Credible actors are identifiable: community cultural authorities can authorize or refuse use; archives can operate and fund projects; and the National Lottery Heritage Fund funds and guides oral-history work. The Library of Congress–Passamaquoddy project demonstrates a real community-governed archival workflow. None of these actors expresses demand for eigenmode-based coverage allocation specifically.","source_ids":["S1","S2","S3","S4"]},"prior_art":{"proximity":"ADJACENT_PRIOR_ART","closest_analogues":[{"name":"Comparative analysis of repeated oral-history tellings","similarity":"Compares the same episode across archived tellings and explicitly models content, discourse, context, and archival situatedness.","remaining_difference":"It provides interpretive comparative methodology rather than a transition operator, modal gain spectrum, held-out reconstruction rule, or sampling allocation.","source_ids":["S5"]},{"name":"Motif-network modeling of narrative variants","similarity":"Represents multiple versions of one tale as a network of motifs and narrative progression, with reconstruction as an objective.","remaining_difference":"It models variants and supports generation; it does not estimate successive-performance modes, govern access, or select future authorized observations.","source_ids":["S6"]},{"name":"Dynamic mode decomposition of textual embeddings","similarity":"Applies a modal spectral method to ordered text states and interprets retained versus missing modes.","remaining_difference":"Its states are sentence embeddings within generated and reference paragraphs, not community-reviewed narratological features across living performances; it does not address archival sampling or cultural authority.","source_ids":["S7"]},{"name":"Community-partnered, gap-oriented oral-history collection practice","similarity":"Archives consult communities, seek collection gaps, document context, use time-coded records, and evaluate projects under permissions and deposit agreements.","remaining_difference":"Existing guidance uses scope, consultation, and professional judgment rather than a spectral coverage metric or comparator-tested stopping rule.","source_ids":["S2","S3","S4"]}],"distinctive_claim_remaining":"Within a rights-holder-approved, demonstrably comparable sequence of performances, a regularized modal model of coupled narratological features will yield stable modes that improve held-out prediction and identify consequential under-coverage; selecting permitted recording opportunities from that map will add community-designated variation or reduce held-out reconstruction error more than equal-count performer/occasion quotas, maximum-difference, clustering, or random authorized sampling, without materially changing performance conditions.","confidence":"MODERATE"},"implementation_evidence":{"support":"WEAK","rationale":"Technical building blocks are credible: time-coded archival descriptions are standard, repeated tellings can be compared, motif networks model variants, and DMD can operate on ordered text vectors. However, the central transition is not yet justified: chronological adjacency between performances does not prove narrative transmission from one state to the next, contexts may dominate, and 18 training recordings provide a fragile basis for a seven-dimensional transition map and eigenspectrum. Workflow feasibility depends on existing digitization, traceable coding, approved coders, secure derived-data handling, and community review. Legal authority is collection-, jurisdiction-, person-, and use-specific; permission to listen does not necessarily authorize feature extraction, derived modal loadings, or cross-record comparison. Safety improves through a read-only first step, residual visibility, prohibitions on authenticity ranking, and a halt rule, but external coordinates, reidentification, restricted residuals, and sampling-induced changes remain unresolved until a named authority approves the exact protocol.","source_ids":["S1","S2","S4","S5","S6","S7"]},"scores":{"meaningful_impact":{"score":3,"rationale":"Avoiding archival gaps and preserving changing community traditions matter, but the incremental impact attributable to modal under-coverage has not been observed.","source_ids":["S3","S4"]},"stakeholder_pull":{"score":2,"rationale":"Archives, communities, and a funder express need for gap-aware, governed collection; no source requests this analytic method.","source_ids":["S2","S3","S4"]},"incremental_advantage":{"score":2,"rationale":"The proposal supplies a falsifiable comparison against simpler sampling methods, but no retrospective or prospective performance advantage exists yet.","source_ids":["S5","S6","S7"]},"distinctiveness_plausibility":{"score":3,"rationale":"Narrative-variant networks, repeated-telling analysis, and text DMD are adjacent, while their combination as a community-governed sampling rule appears contrastive in the reviewed evidence. This is not a world-novelty finding.","source_ids":["S5","S6","S7"]},"technical_implementability":{"score":2,"rationale":"Coding and decomposition are implementable, but transition validity, small-sample spectral stability, contextual confounding, and coder dependence are major unresolved assumptions.","source_ids":["S4","S5","S7"]},"adoption_authority_feasibility":{"score":2,"rationale":"Real community-authorized archival structures exist, but authorization is necessarily local and may not extend to derived analysis or new invitations.","source_ids":["S1","S2","S4"]},"evidence_readiness":{"score":2,"rationale":"A retrospective test can be specified, but it requires proprietary recordings, permission records, approved coders, and community judgments unavailable through web research.","source_ids":["S2","S4","S5"]},"safety_net_benefit":{"score":4,"rationale":"A read-only audit, held-out residuals, explicit scope limits, and rollback to unchanged archive practice provide a strong safety net if faithfully governed.","source_ids":["S1","S2","S4"]},"scalability":{"score":2,"rationale":"Software could be reused, but every narrative requires renewed authority mapping, feature review, contextual validation, coding, and interpretation; these are intentionally non-automatable constraints.","source_ids":["S1","S2","S4","S8"]}},"score_confidence":"LOW","costs":{"first_evidence":{"band_2026_usd":"10K_TO_50K","scope":"Permission and data-governance review; census of at most 24 existing recordings; community-reviewed rubric; two approved coders; time-coded provenance; transition-validity screen; regularized models and comparators; six-recording holdout; aggregate rights-holder review.","confidence":"LOW","assumptions":["Recordings are already digitized, documented, and technically accessible.","Existing agreements permit a read-only feasibility review or can be amended without new field collection.","Approximately 250–500 combined specialist hours cover governance, coding, analysis, and review.","No full new transcription, translation, participant contact, or travel is required.","The band is a resource-equivalent estimate, not a vendor quote."],"source_ids":["S4","S8"]},"initial_deployment_startup":{"band_2026_usd":"50K_TO_250K","scope":"One-archive production setup: expanded authorized corpus, secure derived-data environment, workflow integration, coder training, governance documentation, model validation, reporting templates, privacy review, and staff handoff.","confidence":"LOW","assumptions":["The feasibility study does not falsify transition comparability or modal stability.","One archive and one narrative tradition are in scope.","Community authority and approved experts are compensated.","Existing repository infrastructure is reused.","Costs rise materially if translation, transcription, travel, or legal negotiation is extensive."],"source_ids":["S1","S2","S4","S8"]},"operational_launch":{"band_2026_usd":"50K_TO_250K","scope":"First prospective, rights-holder-approved launch: a bounded set of additional sessions, narrator and community compensation, travel and recording, documentation, secure deposit, re-estimation, comparator evaluation, and community decision review.","confidence":"LOW","assumptions":["Approximately 10–20 additional sessions are authorized.","Each new session has a resource burden analogous to the 30–60 hours cited for complete oral-history interview workflows, with additional governance and analysis.","No sacred, restricted, or otherwise prohibited material is solicited.","A suitable archive already accepts the deposits."],"source_ids":["S1","S4","S8"]},"annual_recurring":{"band_2026_usd":"50K_TO_250K","scope":"Active single-archive program covering periodic approved sessions, coder refresh and quality checks, permissions maintenance, spectral and residual re-estimation, secure storage, community review, and evaluation against simpler sampling.","confidence":"LOW","assumptions":["The archive conducts roughly 5–15 approved sessions or equivalent reviews annually.","Professional staff, community reviewers, narrators, translators where needed, and data stewards receive fair compensation.","Recurring software expense is minor relative to governance, documentation, and human review.","A dormant read-only monitoring program could fall into the 10K_TO_50K band."],"source_ids":["S4","S8"]}},"verified_pipeline_gates":{"externally_supported_problem":{"status":"UNCERTAIN","reason":"The importance of archival gaps and context-sensitive variation is supported, but the asserted mode-specific sampling failure is not externally observed or measured.","source_ids":["S3","S4","S5"]},"externally_credible_adopter_or_authorizer":{"status":"YES","reason":"Archives and heritage funders are credible operators and funders, while the Library of Congress–Passamaquoddy example verifies a community cultural authority with real control over archival representation and reuse.","source_ids":["S1","S2","S3","S4"]},"distinct_testable_incremental_claim":{"status":"YES","reason":"The claim names outcomes, comparators, conditions, and failure criteria: held-out prediction and consequential-variation coverage must exceed quota, maximum-difference, clustering, context-only, and random authorized approaches.","source_ids":["S5","S6","S7"]},"bounded_next_evidence_step":{"status":"YES","reason":"A read-only, at-most-24-recording feasibility study with approved coding, a fixed holdout, explicit comparators, and stop conditions is bounded, although it cannot establish prospective intervention efficacy.","source_ids":["S4","S5","S7"]},"no_unresolved_safety_or_authority_stop":{"status":"UNCERTAIN","reason":"No named collection's agreements were reviewed. Listening rights may not cover feature extraction, cross-performance comparison, derived loadings, residual examples, or later invitations, and the appropriate community authority may be plural.","source_ids":["S1","S2","S4"]},"credible_cost_scope_and_range":{"status":"YES","reason":"All four estimates state their scope and assumptions and are anchored to official workflow and labor-burden guidance, though local compensation, translation, travel, and governance costs keep confidence low.","source_ids":["S4","S8"]}},"next_evidence_step":"With no new recording, a named archive and community authority should jointly pre-register a feasibility study on no more than 24 already authorized recordings. First, audit whether permissions cover feature extraction, cross-record comparison, derived data, approved reviewers, retention, and deletion. Then determine whether at least 18 training and six held-out performances form genuine, context-justified transmission or repeated-telling sequences; chronological ordering alone is insufficient. Two approved coders should apply a community-reviewed, time-coded rubric, with disagreements adjudicated without exposing restricted passages. Compare a regularized transition model and its modal reduction against persistence, context-only and narrator/occasion models, raw-feature ridge regression, PCA plus regression, clustering, quota, maximum-difference, and random authorized selection using narrator- and context-blocked validation. Pre-register falsifiers: inadequate authority; no defensible transition relation; unacceptable coder agreement; modes unstable under coder, bootstrap, ordering, and context perturbations; no reproducible retained/omitted separation; modal prediction not better than the strongest simpler comparator; consequential information concentrated in residuals; reidentification risk; or rights-holder rejection of the representation. Treat this as a feasibility screen, not evidence that mode-guided new collection works. A subsequent prospective comparison of authorized invitations would be required for the intervention claim.","blocking_evidence":["No collection-specific permission or authority documents show that the proposed derived analysis is authorized.","No external evidence establishes the prevalence or consequence of mode-specific under-coverage in living-narrative archives.","The existence, number, and independence of comparable performance-to-performance transitions are unknown.","The seven-feature rubric's cultural acceptability, coder reliability, and resistance to contextual confounding are untested.","Modal stability, spectral separation, and held-out advantage over simpler models require proprietary recordings and live analysis.","Prospective improvement from mode-guided invitations, including whether invitations alter performance conditions, requires an authorized field comparison.","Community-designated consequential variations and acceptable residual disclosure rules cannot be determined from public sources."],"research_disposition":"PARTNERED_RESEARCH_PROGRAM","world_novelty_boundary":"The eight-source review found adjacent repeated-telling comparison, motif-network narrative modeling, text dynamic-mode decomposition, and community-governed archival practice, but no direct demonstration of spectral mode-coverage sampling for living narrative archives. This bounded search cannot measure world novelty, patentability, freedom to operate, market size, realized impact, or the full unpublished and non-English practice landscape.","arm":"COMPLETE_PROPOSAL_PORTFOLIO","candidate_version":0,"controller_recommendation":{"action":"STOP_EMPIRICAL_RESEARCH_NEEDED","repairable":false,"material_progress_observed":false,"progress_targets":["Obtain project-specific written authorization covering coding, cross-record comparison, derived modal data, residual review, access roles, retention, and deletion.","Document a defensible transmission or repeated-telling relation and sufficient authorized transition count; stop if recordings are merely chronologically adjacent independent performances.","Secure community approval of the state vector, context boundaries, consequential-variation criteria, compensation, privacy controls, and interpretation contract.","Demonstrate acceptable two-coder reliability and stability under coder, bootstrap, ordering, narrator, and context perturbations.","Pre-register leakage-resistant held-out comparisons against persistence, context-only, raw-feature, PCA, clustering, quota, maximum-difference, and random authorized baselines.","Show reproducible held-out improvement and residual safety before requesting any new performance.","If retrospective feasibility passes, conduct a separately authorized prospective comparison showing that mode-guided invitations add consequential coverage without measurably changing the practice being observed."],"reason":"Web research supports the broad archival need, governance structure, and adjacent methods but cannot resolve the candidate's decisive claims. They require collection-specific permissions, proprietary recordings, approved human coding, community judgment, and ultimately a live prospective sampling comparison. Under the required controller rule, that evidence dependency mandates STOP_EMPIRICAL_RESEARCH_NEEDED; all STOP actions are non-repairable."},"proposal_index":5}