{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp05_complete_proposal_portfolio20_20260803","cell_id":"negative_space_design__astronomy_astrophysics","arm":"COMPLETE_PROPOSAL_PORTFOLIO","candidate_id":"nsd-astro-annotation-free-morphology-moat","proposal_index":2,"version":0,"title":"Annotation-Free Moats for Faint-Structure Review","problem":"Astronomers vetting faint extended structures in survey cutouts may view segmentation outlines, source symbols, labels, crosshairs, mask boundaries, and catalog footprints over the same pixels whose morphology they must judge. When these overlays enter the candidate footprint, their edges can obscure low-contrast structure or supply a visual shape that is not present in the measurements. The reviewer can no longer cleanly separate pixel evidence from interface annotation.","actors":["Astronomers reviewing survey candidates","Survey visualization and pipeline developers","Catalog and segmentation-pipeline maintainers","Scientists responsible for candidate classification or follow-up selection","Data-quality specialists responsible for masks and artifact flags"],"observable_state":"For a bounded set of archived candidate cutouts, the interface can record which overlay layers intersect the candidate footprint, the area and boundaries covered by annotation marks, whether reviewers toggle layers before deciding, the morphologies and artifact explanations they report from each view, their decision time and confidence, and errors caused by missing quality or positional context.","consequence":"A candidate may be dismissed, promoted, or morphologically classified on the basis of marks that conceal or visually organize the underlying pixels, weakening the traceability of subsequent catalog or follow-up decisions.","affected_objective":"Enable reviewers to distinguish faint astrophysical structure from processing artifacts and neighboring-source contamination while retaining the provenance, masks, coordinates, and catalog context needed to interpret the image correctly.","intervention":"Add a protected annotation-free moat to the candidate-review interface. The moat is a bounded region derived from the candidate footprint plus a declared margin. Its default focal view contains the calibrated image pixels but excludes nonessential crosshairs, labels, catalog symbols, segmentation outlines, and footprint strokes from that region. Context outside the moat remains visible, and an adjacent status frame identifies every omitted layer and preserves scale, orientation, band, stretch, provenance, and data-quality state. Safety-relevant invalid-pixel, saturation, detector-edge, and unavailable-data warnings are exempt from hiding. A single documented press-and-hold or paired-view control restores all omitted overlays without changing the zoom or stretch. The moat is protected by a layer-admission rule so newly added annotations cannot enter it by default without a perception and context check.","structural_mapping":[{"archetype_element":"Attention Competition Map","domain_realization":"Inventory every mark competing with faint pixel structure and distinguish contextual annotations from quality warnings whose visibility is necessary."},{"archetype_element":"Omission Candidate","domain_realization":"Treat crosshairs, catalog symbols, labels, segmentation strokes, and noncritical footprint boundaries intersecting the focal region as candidates for reversible omission."},{"archetype_element":"Protected Empty Space","domain_realization":"Reserve a candidate-centered region in which nonessential annotation marks cannot appear in the default morphology view."},{"archetype_element":"Positive Form Relationship","domain_realization":"The absence of marks is tied specifically to making the measured extent, asymmetry, discontinuity, and relation to the local background perceptible without an imposed outline."},{"archetype_element":"Absence Boundary","domain_realization":"Define the moat from a recorded candidate footprint and margin rule, with explicit exemptions for critical data-quality information."},{"archetype_element":"Context Preservation Frame","domain_realization":"Keep coordinates, scale, orientation, band, stretch, provenance, omitted-layer inventory, and surrounding sources visible outside or adjacent to the moat."},{"archetype_element":"Meaning-of-Absence Check","domain_realization":"Label the quiet region as an annotation-suppressed view so reviewers do not confuse absent marks with missing catalogs, failed loading, unavailable masks, or unprocessed data."},{"archetype_element":"Reintroduction Trigger","domain_realization":"Restore all annotations through one obvious control or synchronized paired view while preserving the exact pixel framing for comparison."},{"archetype_element":"Accessibility and Recoverability Guardrail","domain_realization":"Provide a persistent layer inventory, keyboard-operable restoration, non-color-only status cues, and exemptions for saturation, invalid-pixel, detector-edge, and missing-data warnings."},{"archetype_element":"Clarity or Effect Test","domain_realization":"Compare what reviewers perceive and decide from the protected view, the fully annotated baseline, and a rival paired-view design using frozen cutouts and recorded interpretations."}],"mechanism_mapping":[{"mechanism_slug":"focus_mode_or_control_hiding","role":"Temporarily clears secondary visual controls and overlays from the focal evidence while guaranteeing immediate restoration and exempting critical warnings.","counterfactual_removal":"Without this mechanism, the intervention would either leave competing marks in place or remove them without a reliable way for reviewers to recover needed context."},{"mechanism_slug":"editorial_cut","role":"Classifies annotation layers by whether their removal sharpens pixel interpretation without destroying load-bearing context, and records every reversible omission.","counterfactual_removal":"Without this mechanism, the moat could become indiscriminate minimalism that hides masks, provenance, or neighboring-source relationships."},{"mechanism_slug":"margin_and_gutter_system","role":"Turns the quiet area into a repeatable footprint-and-margin rule that resists encroachment by newly added interface layers across many cutouts.","counterfactual_removal":"Without this mechanism, annotation clearance would be adjusted inconsistently case by case and would erode as new overlays were introduced."},{"mechanism_slug":"whitespace","role":"Uses the absence of annotation strokes around the candidate to separate measured morphology from interface-supplied shapes and grouping cues.","counterfactual_removal":"Without this mechanism, reviewers would still have to perceptually disentangle the candidate's pixel structure from marks occupying the same region."}],"causal_chain":["Overlay strokes and symbols are mapped against the candidate footprint to identify which marks compete directly with faint pixel evidence.","Nonessential intersecting annotations are reversibly omitted while critical quality warnings and interpretive context are preserved.","The bounded moat prevents interface marks from imposing contours or covering low-contrast structure in the default focal view.","Reviewers first inspect morphology in the resulting annotation absence rather than through catalog or segmentation geometry.","The explicit layer inventory prevents the absence from being mistaken for missing data or failed processing.","Immediate overlay restoration lets reviewers test the pixel-based interpretation against positions, masks, segmentations, and catalog associations.","Recorded disagreements between the quiet and annotated views reveal whether annotations were clarifying evidence, obscuring it, or supplying an unsupported form."],"baseline":"Candidate-review cutouts display the calibrated image together with all enabled catalog, segmentation, mask, footprint, label, and crosshair layers; reviewers may manually toggle individual layers when they notice interference.","nearest_rivals":["A conventional layer panel that lets each reviewer manually toggle annotations","Automated label-placement and collision-avoidance that moves marks away from dense regions","Synchronized side-by-side raw and fully annotated cutouts without a protected candidate-centered moat"],"remaining_contrastive_claim":"The proposal does not merely offer optional layer controls or rearrange labels. It establishes a default, bounded absence of nonessential marks exactly where morphology is judged, protects that absence against layer creep, preserves an explicit account of what is missing, and makes restoration immediate. Testing must determine whether that structural separation adds value beyond ordinary toggles or paired images.","authority_safety":{"decision_authority":"A survey visualization lead may authorize a read-only prototype and evaluation, with the data-quality lead approving which warning layers are never suppressible. Catalog, classification, and observing authorities retain all scientific decision rights.","authorized_first_step":"Render duplicate, nonoperational review interfaces for a bounded archived-cutout evaluation; do not alter source pixels, production catalogs, trained classifiers, live candidate states, or observing priorities.","excluded_actions":["Changing calibrated pixel values or image-processing parameters between compared views","Suppressing saturation, invalid-pixel, detector-edge, unavailable-data, or other approved critical warnings","Editing production classifications or catalogs from the evaluation interface","Using prototype decisions to allocate follow-up observations","Making annotations unrecoverable or requiring an undocumented reveal gesture","Allowing the moat boundary to imply a measured astrophysical extent"],"halt_rollback":"Stop if any view changes the underlying pixel rendering, hides an approved critical warning, fails to disclose an omitted layer, or cannot restore overlays in exact registration. Rollback consists of disabling the prototype and retaining the unchanged archived images, annotations, and production interface."},"negative_tests":{"strongest_counterevidence":"Reviewers using fully annotated or ordinary paired views identify artifacts, source associations, and morphology at least as consistently and efficiently as reviewers using the moat, while the moat causes missed mask conditions, mistaken continuity, or framing-induced interpretations.","problem_falsifier":"The problem is falsified for the tested workflow if nonessential overlay marks rarely intersect the morphology-review region or if blinded reviewers' descriptions and classifications do not materially depend on whether those marks are present.","intervention_falsifier":"The intervention is falsified if the protected view does not improve separation of pixel-based observations from annotation-derived interpretations relative to the strongest rival, or if preserving enough context to avoid quality and attribution errors eliminates the proposed perceptual benefit.","risks":["A hidden or poorly surfaced mask could make discontinuous valid data appear continuous.","The moat boundary could anchor reviewers to a pipeline-supplied candidate extent.","Suppressing neighboring-source annotations could weaken contamination judgments.","Reviewers might interpret the quiet default as institutionally preferred evidence and discount useful catalog context.","The restoration control may be undiscoverable or burdensome for keyboard, screen-magnification, or alternative-input users.","A fixed margin rule may be unsuitable across source scales, point-spread functions, bands, and display sizes.","Maintaining two perceptual states may increase review time or create registration mistakes."]},"next_evidence_step":"Choose no more than 60 archived cutouts spanning prespecified overlay densities, candidate angular extents, and known quality-mask conditions. Freeze pixels, stretch, zoom, and all annotations. In a randomized crossover, have qualified reviewers examine the fully annotated baseline, the protected-moat prototype, and the strongest feasible rival. Before revealing reference information, record free-form morphology descriptions, traced boundaries, artifact and contamination explanations, classification choices, confidence, time, overlay restorations, and missed quality warnings. Separately audit pixel identity, registration, accessibility, and critical-warning visibility. Use the results only to decide whether a prospective evaluation is warranted.","prior_art_status":"UNSEARCHED","diversity_from_prior_proposals":"Proposal 1 addressed operational scarcity in time-domain observing by reserving genuinely unassigned telescope time near forecast information updates, causing late evidence to encounter schedulable capacity. This proposal addresses perceptual interference during scientific image review by reserving an annotation-free spatial region over unchanged pixel data, causing morphology to be inspected before interface marks impose or conceal form. It changes the affected object from an observing schedule to a review image, the actors from scheduling and operations authorities to image reviewers and visualization maintainers, the intervention from temporal capacity reservation to reversible overlay omission, the causal path from reduced replanning conflict to separation of measurements from graphical cues, and the adoption unit from observatory scheduling policy to a standalone review-interface protocol. It neither extends nor implements the response-window proposal.","revision_record":{"parent_version":null,"progress_targets_addressed":["Second independently adoptable proposal","Material problem and intervention diversity from proposal 1","Causal fidelity to protected absence","Operational authority and safeguards","Contrastive rivals and bounded evidence"],"conceptual_changes":[],"operational_changes":[],"evidence_changes":[],"claim_changes":[]}}