{"actors":["Digital literary corpus curator","Repository engineer responsible for ingestion and migration","Textual bibliographers who identify form-bearing features","Literary scholars who use the corpus for prosodic, rhetorical, and genetic analysis","Data steward responsible for preservation records"],"affected_objective":"Preserve the recoverability of form-bearing textual features—lineation, stanza boundaries, indentation, punctuation, typography, and page-linked paratext—in a digitized poetry corpus through repeated transformations.","arm":"ORDINARY_DIVERSE_P2","authority_safety":{"authorized_first_step":"The corpus curator may authorize a read-only assessment of sentinel passages and a stratified sample from one corpus release; assessors may compare derivatives with designated reference images and record suspected losses but may not modify public files or metadata.","decision_authority":"The corpus curator assigns ordinary reprocessing and metadata repairs; the data steward approves replacement of preservation derivatives, and the repository engineer controls release rollback or migration suspension when an unacceptable threshold is confirmed.","excluded_actions":["Treating every difference from page images as deterioration without accounting for documented editorial policy","Silently replacing public corpus files","Inferring authorial intention from layout signals alone","Using condition results to evaluate individual transcribers or scholars","Discarding prior derivatives or provenance records after repair","Suspending access to unaffected corpus segments"],"halt_rollback":"Halt the assessment if reference images or transformation histories cannot support reliable comparison, if sampling risks exposing restricted materials, or if condition labels are applied inconsistently. Remove pilot-only labels, leave public derivatives unchanged, and retain the access-controlled assessment record for rubric revision."},"baseline":"At pilot start, each sampled item is linked to a designated reference image or preservation transcription and to its transformation history. Healthy means all designated form-bearing features remain representable and correctly ordered; acceptable means isolated, reversible display or metadata defects do not alter feature recovery; warning means repeated loss, merger, displacement, or ambiguity affects analysis but can still be reconstructed from retained sources; unacceptable means a necessary feature cannot be recovered reliably from the corpus derivative and its available provenance. Persistence requires the same failure class in two inspections or across two transformation stages; acceleration means the affected share or severity rises over successive releases.","candidate_id":"deterioration_monitoring__literature_literary_theory__ORDINARY_DIVERSE_P2","causal_chain":["Poetry files pass through OCR correction, schema conversion, Unicode normalization, interface redesign, and repository migration.","Each transformation can erase or reinterpret small form-bearing distinctions such as indentation, stanza breaks, emphasis, punctuation, or page association.","The derivatives remain readable and searchable, so users can accommodate local anomalies without recognizing cumulative loss of formal evidence.","Recurring comparisons of sentinel passages and stratified samples against designated references reveal feature-specific mismatch, ambiguity, and provenance-gap signals.","Release-by-release histories distinguish isolated rendering errors from persistent or accelerating loss concentrated in particular formats, periods, or transformation stages.","Warning and unacceptable bands activate named reprocessing, migration-control, and preservation authority.","Curators restore mappings, regenerate derivatives, quarantine a faulty transformation, or publish a scoped limitation while reference material still permits reconstruction.","Comparison of flags with bibliographic adjudication and repair outcomes recalibrates the sample, feature rubric, thresholds, and inspection cadence."],"cell_id":"deterioration_monitoring__literature_literary_theory","consequence":"A corpus may remain operational while gradually becoming unsuitable for analyses that depend on poetic form, causing researchers to treat transformation artifacts as textual evidence and making later reconstruction harder when provenance or reference files are no longer readily connected.","diversity_from_prior_proposals":"This opportunity concerns degradation of encoded formal and paratextual features in a poetry corpus, addressed through reference-based sentinel testing and transformation control. Its affected object, operational intervention, responsible actors, and causal path are separate from monitoring the evidentiary warrant of interpretive commentary in a living critical edition.","experiment_id":"eoa_inverse_innovation_exp13_second_slot_policy60_20260806","intervention":"Install a Poetic-Form Fidelity Watch for a digitized poetry corpus. Before each derivative release or migration, and quarterly between releases, the curator runs fixed sentinel passages plus a stratified sample through feature-level comparison with designated reference images or preservation transcriptions. Reviewers separately record line and stanza boundaries, indentation, punctuation, emphasis, character identity, reading order, and page-linked paratext, preserving results by file version and transformation stage. One severe mismatch prompts manual verification; persistence of a feature-loss class across two inspections requires an owned reprocessing plan; confirmed unrecoverability or systematic corruption pauses the affected transformation and triggers regeneration, rollback, source reacquisition, or a scoped limitation notice. After adjudication, the curator revises noisy sentinels, sampling strata, thresholds, or cadence.","mechanism_mapping":[{"counterfactual_removal":"Without recurring reference comparisons, a derivative can remain readable while successive transformations silently remove evidence needed for formal analysis.","mechanism_slug":"preventive_inspection","role":"Checks sentinel passages and sampled items before transformation defects become embedded in later corpus releases."},{"counterfactual_removal":"Without feature-specific histories, curators cannot tell whether a mismatch is isolated, persistent, spreading, or associated with a particular processing stage.","mechanism_slug":"quality_drift_monitoring","role":"Tracks longitudinal drift in the fidelity of lineation, typography, punctuation, reading order, and paratext."},{"counterfactual_removal":"Without explicit tracking of unresolved format and provenance defects, warnings can remain detached from reprocessing work.","mechanism_slug":"maintenance_backlog_review","role":"Connects threshold crossings to an owner, affected files, transformation stage, response, and due point."},{"counterfactual_removal":"Without comparison to designated page images or preservation transcriptions, the monitored signals lack a trustworthy condition reference.","mechanism_slug":"data_integrity","role":"Maintains reference identity, version lineage, and auditable comparison inputs."},{"counterfactual_removal":"Without bibliographic examination of the material text, automated differences could confuse intentional encoding policy with loss of meaningful form.","mechanism_slug":"close_reading","role":"Supplies domain-specific adjudication of whether a flagged visual or textual distinction is analytically load-bearing."}],"nearest_rivals":["Checksum validation detects unintended bit changes but not semantically damaging transformations that produce internally valid new files.","A one-time digitization quality audit assesses an ingestion event but does not preserve recurring trend memory across later conversions and releases.","Schema validation checks structural conformance without establishing whether encoded structure still represents the source's form-bearing distinctions.","Routine repository backups preserve earlier files but do not identify when current derivatives have crossed a condition threshold or require rollback.","Scholarly collation identifies textual variants but does not by itself create a recurring condition, escalation, and corpus-repair loop."],"negative_tests":{"intervention_falsifier":"The intervention is undermined if sentinel and sampled signals do not precede independently adjudicated loss, if different bibliographers cannot apply the feature rubric consistently, or if confirmed threshold crossings cannot be connected to feasible reprocessing, rollback, or limitation notices.","problem_falsifier":"The inferred problem is unsupported if comparison across transformation histories shows no directional or persistent loss of form-bearing features, observed mismatches are confined to random display incidents, or existing release controls already detect and repair the relevant changes before publication.","risks":["Reference images may themselves be incomplete or ambiguous.","A fixed sentinel set may miss deterioration outside familiar formats or canonical authors.","Automated comparisons may overflag legitimate editorial normalization.","Aggregate rates may conceal severe loss in a small corpus segment.","Frequent manual adjudication may consume preservation capacity.","Repair may improve visual similarity while damaging accessibility or machine readability.","Limitation notices may be misunderstood as judgments about the literary value of affected works."],"strongest_counterevidence":"Some typographic and layout differences are accidental products of particular printings rather than stable textual features, so closer visual correspondence to a reference copy does not necessarily produce a more defensible scholarly representation."},"next_evidence_step":"Conduct a read-only assessment on one existing poetry-corpus release using 30 preselected sentinel passages and a stratified sample of 70 additional items spanning file formats, publication periods, and layout complexity. Two textual bibliographers independently compare current derivatives with designated references without seeing transformation-stage flags. Reconstruct results for the two most recent processing stages, then record agreement, false-alarm classes, feature-loss direction and persistence, time required, provenance sufficiency, and whether each proposed threshold maps to existing curator or repository authority. Do not alter or republish corpus files during the assessment.","observable_state":"A versioned condition ledger records each sampled item and reference, transformation stages, line and stanza boundary mismatches, indentation displacement, punctuation or character substitutions, lost emphasis, reading-order defects, detached paratext, provenance gaps, signal direction and persistence, adjudicated severity, affected corpus segment, assigned owner, response status, and time since threshold crossing.","prior_art_status":"UNSEARCHED","problem":"A digitized poetry corpus can gradually lose analytically significant formal features as files are OCR-corrected, normalized, converted, and migrated. Small changes to lineation, indentation, typography, punctuation, reading order, or page-linked paratext often leave works readable and searchable, so each loss appears tolerable. Without recurring, feature-level condition records, successive derivatives can accumulate distortions that compromise prosodic, rhetorical, or genetic analysis before curators recognize a corpus-level failure.","proposal_index":2,"remaining_contrastive_claim":"The defining claim is that recurring, reference-based tracking of feature-specific loss across transformation stages can activate scoped reprocessing or rollback before a still-readable poetry corpus ceases to preserve evidence required for formal literary analysis.","revision_record":{"claim_changes":["Initial formulation; no parent-version claims were revised."],"conceptual_changes":["Initial transfer of deterioration monitoring to preservation of form-bearing features in transformed digital poetry corpora."],"evidence_changes":["No external evidence was consulted; prior art remains unsearched."],"operational_changes":["Initial specification of reference-linked condition bands, sentinel and stratified inspections, persistence thresholds, transformation controls, authority boundaries, and a read-only assessment."],"parent_version":null,"progress_targets_addressed":["Materially independent affected objective","Observable feature-level deterioration signals","Transformation-stage trend memory","Threshold-linked corpus response","Authority safeguards and rollback","Bounded falsification step"]},"schema_version":1,"structural_mapping":[{"archetype_element":"Condition to preserve","domain_realization":"Recoverability of form-bearing textual and paratextual features in digital poetry derivatives."},{"archetype_element":"Baseline and condition bands","domain_realization":"Healthy, acceptable, warning, and unacceptable states defined through designated references, documented editorial policy, recoverability, and analytical consequence."},{"archetype_element":"Leading and lagging deterioration indicators","domain_realization":"Leading signals include transformation-stage mismatches, provenance gaps, and growing feature ambiguity; lagging signals are bibliographic judgments that necessary formal evidence is no longer reliably recoverable."},{"archetype_element":"Inspection cadence","domain_realization":"Sentinel and stratified inspection before each derivative release or migration, plus quarterly review between releases."},{"archetype_element":"Trend interpretation","domain_realization":"The ledger tracks direction, persistence across inspections or stages, acceleration across releases, and concentration by format, period, or transformation process."},{"archetype_element":"Thresholds and escalation","domain_realization":"A severe mismatch triggers verification, persistent feature loss triggers an owned reprocessing plan, and systematic or unrecoverable corruption pauses the affected transformation for steward review."},{"archetype_element":"Response pathway","domain_realization":"Correct mappings, regenerate derivatives, quarantine a faulty transformation, roll back the affected release component, reacquire a source, or publish a scoped limitation notice."},{"archetype_element":"Recalibration","domain_realization":"Compare monitoring flags with independent bibliographic adjudication and repair results, then revise sentinels, strata, feature definitions, thresholds, or cadence."}],"title":"Poetic-Form Fidelity Watch for Transforming Digital Corpora","version":0}