{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp05_complete_proposal_portfolio20_20260803","cell_id":"negative_space_design__marine_science","arm":"COMPLETE_PROPOSAL_PORTFOLIO","candidate_id":"nsd-ms-cause-coded-time-series-gaps","proposal_index":3,"version":0,"title":"Cause-Coded Gaps in Ocean Time-Series Displays","problem":"An ocean-observing dashboard can visually connect valid measurements across periods when a sensor was offline, communications failed, values were rejected by quality control, or processing remained incomplete. The resulting continuous line fills an interval in which no accepted observation exists. Viewers may then read apparent stability, gradual change, or event timing from display geometry generated by the interface rather than from measurements.","actors":["Ocean-observing program data manager","Instrument and calibration scientist","Quality-control analyst","Data-portal designer","Marine scientist using the time series","Operational user relying on current sensor status"],"observable_state":"A primary time-series plot draws lines or filled areas between accepted samples on opposite sides of an unobserved interval; missingness causes are available only in flags, tables, or hover text; visually similar gaps represent sensor outage, rejected values, delayed processing, and true zero measurements; users can describe conditions inside an interval despite there being no accepted observations there.","consequence":"Users may mistake an epistemically unknown interval for observed continuity, assign unsupported timing or shape to an ocean event, or confuse absence of accepted data with an observed zero or stable condition.","affected_objective":"Make ocean time-series products interpretable without implying measurements or continuity where accepted observations are absent, while retaining access to modeled estimates and operational status when those are needed.","intervention":"In a bounded data product, stop the primary observed-data trace at the last accepted sample before each qualifying interval and resume it at the first accepted sample afterward. Protect the intervening visual gap from automatic line connection, area fill, or default interpolation. Bound the gap with endpoint marks and a minimal cause code distinguishing no sample, quality-control rejection, instrument offline, communications outage, and processing pending. Keep surrounding axes, units, station identity, sampling cadence, and quality context visible so the gap remains oriented. If an estimated bridge is scientifically required, place it in an explicitly activated secondary layer with distinct styling, method identity, and uncertainty representation; never merge it into the observed trace. Replace the gap only when accepted data arrive or a formally labeled estimate layer is requested.","structural_mapping":[{"archetype_element":"Attention Competition Map","domain_realization":"The design identifies which marks compete to define the interval: accepted samples, automatic connecting lines, filled areas, imputed estimates, quality flags, alarms, and neighboring variables."},{"archetype_element":"Omission Candidate","domain_realization":"Synthetic line segments and area fills spanning intervals without accepted observations are removed from the primary observed-data layer."},{"archetype_element":"Protected Empty Space","domain_realization":"The horizontal interval between the last and next accepted samples remains visually unfilled and cannot be consumed by default interpolation."},{"archetype_element":"Positive Form Relationship","domain_realization":"The gap separates and frames the two observed segments, making their evidentiary boundary visible rather than depicting a fabricated transition between them."},{"archetype_element":"Absence Boundary","domain_realization":"Endpoint marks, timestamps, and the missingness interval define exactly where accepted observation stops and resumes."},{"archetype_element":"Meaning-of-Absence Check","domain_realization":"A cause code distinguishes unobserved, rejected, offline, delayed, and zero-valued states so blankness is not interpreted generically."},{"archetype_element":"Context Preservation Frame","domain_realization":"Axes, units, station identity, cadence, adjacent accepted measurements, and relevant operational status remain visible around the gap."},{"archetype_element":"Reintroduction Trigger","domain_realization":"The observed trace resumes only when the pipeline accepts a measurement; a modeled bridge appears only through an explicit user action and remains separately identified."},{"archetype_element":"Accessibility and Recoverability Guardrail","domain_realization":"Textual interval summaries and cause codes accompany visual gaps, while users can recover raw flags and any approved estimate without relying on color, hover, or gap perception alone."},{"archetype_element":"Clarity or Effect Test","domain_realization":"Users interpret matched baseline and cause-coded-gap displays without knowing the rendering condition, allowing comparison of whether they distinguish observed values, unknown intervals, true zeros, and estimates correctly."}],"mechanism_mapping":[{"mechanism_slug":"empty_state_design","role":"Diagnoses the type of missing interval before rendering it and gives each state an appropriate cause label and exit condition.","counterfactual_removal":"Without state diagnosis, every blank interval would communicate the same generic absence, leaving users unable to distinguish outage, rejection, delay, and intentional lack of sampling."},{"mechanism_slug":"editorial_cut","role":"Removes the unsupported connecting segment while preserving load-bearing context and keeping any approved estimate recoverable in a separate layer.","counterfactual_removal":"Without the editorial-cut discipline, the interface might either retain a misleading bridge or delete surrounding context and modeling access needed for legitimate analysis."},{"mechanism_slug":"whitespace","role":"Uses the unfilled interval itself to separate observed segments and signal that they should not be perceptually grouped as one continuous measured trajectory.","counterfactual_removal":"Without the perceptual gap, endpoint labels and quality flags would compete with a dominant continuous line that still implies connection at a glance."},{"mechanism_slug":"margin_and_gutter_system","role":"Applies a consistent protected-gap rule across stacked variables, stations, zoom levels, and exports so rendering changes do not close the interval selectively.","counterfactual_removal":"Without a reusable spacing and boundary rule, some views could preserve the gap while others reconnect it, producing inconsistent evidentiary meanings."}],"causal_chain":["A sensor or processing pipeline produces an interval without accepted observations.","The baseline renderer connects accepted endpoints or fills the interval despite that evidentiary absence.","The continuous mark becomes the most perceptually salient account of conditions inside the interval.","The intervention removes the unsupported mark and protects the resulting gap across views.","Endpoint boundaries and a cause code explain what is absent and why, while surrounding scientific context remains available.","Users can distinguish measured segments from the unknown interval before deciding whether to inspect a separately labeled estimate.","Interpretations of continuity, event timing, or zero conditions are therefore tied to the status of the data rather than to an automatically completed visual shape."],"baseline":"Plot all accepted samples as a continuous trace by connecting consecutive valid points, optionally using default interpolation or area fill. Expose missingness through metadata flags, a separate status panel, or hover text while leaving the primary visual trajectory intact.","nearest_rivals":["Draw a dashed interpolated bridge directly in the primary trace; this acknowledges estimation but still supplies a visually continuous trajectory before the user chooses to consider a model.","Retain the continuous line and add missing-data symbols or shaded warnings; this preserves compactness but forces annotations to compete with the line's continuity cue.","Show quality flags only in a table or downloadable record; this serves expert audit but does not correct the initial perceptual claim made by the chart.","Impute all missing values with a formal model and display uncertainty; this may support analyses requiring complete series but answers an estimation question instead of preserving the distinction between observation and non-observation.","Break the line without explaining the gap; this removes false continuity but leaves users unable to distinguish deliberate absence from a rendering failure or data-service error."],"remaining_contrastive_claim":"The proposal tests whether protecting a cause-coded visual discontinuity in the primary observed-data layer makes the boundary of observation more interpretable than adding warnings to an otherwise continuous trace, while leaving estimation available as a separate recoverable act.","authority_safety":{"decision_authority":"The observing program's data-product owner may authorize the experimental rendering. The quality-control lead retains authority over whether a value is accepted, rejected, or pending. Operational supervisors retain authority over alarm and safety displays and may exclude them from the trial.","authorized_first_step":"Create a non-production comparison view for one archived mooring variable containing known examples of accepted data, true zero values, sensor outage, quality-control rejection, communications loss, and processing delay; do not alter the underlying records or the operational dashboard.","excluded_actions":["Changing raw measurements, quality flags, or acceptance decisions to create cleaner gaps","Suppressing valid observations or unfavorable measurements","Using blankness to conceal uncertainty, instrument performance, or data provenance","Removing alarms, last-known-status indicators, or safety-critical operational information","Presenting imputed or modeled values as accepted observations","Applying the rendering to production or exported scientific products before the bounded interpretation test is reviewed"],"halt_rollback":"If the trial obscures current operational status, fails accessibility checks, mislabels interval causes, or prevents users from retrieving required estimates, withdraw the comparison view and restore the existing renderer. Preserve the test configuration and findings, correct no source records, and do not propagate the gap rule to other products."},"negative_tests":{"strongest_counterevidence":"In blinded tasks, users interpret baseline connected traces as cautiously as cause-coded gaps because sample markers, cadence, and existing quality indicators already make the lack of observations clear; the gap adds no interpretive distinction and may fragment legitimate reading of the series.","problem_falsifier":"The problem is falsified for the tested product if the primary trace never spans intervals without accepted observations, missingness causes are already perceptually explicit and accessible, and users reliably distinguish unknown intervals from observed zeros and modeled values.","intervention_falsifier":"The intervention is falsified if users remain unable to identify what the gap means, infer unsupported event timing at the same rate as with the baseline, or lose access to context required for legitimate continuous estimates despite the recovery layer.","risks":["Users interpreting a protected gap as a chart failure","Different missingness causes becoming visually indistinguishable at small display sizes","Frequent gaps fragmenting long records and obscuring large-scale patterns","Cause codes overwhelming the chart when several variables fail together","Endpoint placement implying more temporal precision than the sampling cadence supports","Accessible text and exported figures diverging from the interactive display","A separately activated estimate layer being mistaken for observed data","Operational users overlooking a current outage if status information is simplified excessively"]},"next_evidence_step":"Using one archived mooring series, preregister the qualifying gap rules, cause taxonomy, endpoint behavior, estimate-layer styling, accessibility representation, and interpretation questions. Produce matched baseline and intervention views without changing source data. Randomize blinded users to identify which intervals contain accepted observations, true zeros, unknown conditions, rejected values, and modeled estimates, and ask what event timing can be supported. Separately audit rendering across zoom levels, static exports, screen-reader text, and stacked variables. Use the results only to decide whether a limited production pilot is justified.","prior_art_status":"UNSEARCHED","diversity_from_prior_proposals":"Proposal 1 created physical time intervals without research transmissions during data acquisition so passive acoustic signals would not be masked. Proposal 2 created persistent geofenced seafloor areas without contact operations so benthic reference plots would not be disturbed. This proposal changes neither transmission schedules nor physical sampling footprints: it addresses a post-acquisition interpretive error caused when software visually fills intervals lacking accepted ocean observations. Its protected absence is an epistemic gap in a data display; its intervention is a cause-aware rendering and recovery rule; and its causal path runs from unsupported graphic continuity to mistaken inference, then from deliberate visual discontinuity to explicit recognition of observation boundaries. It is independently adoptable by an ocean-data program that performs neither active-acoustic surveys nor benthic contact sampling.","revision_record":{"parent_version":null,"progress_targets_addressed":["Generate one independently adoptable continuation candidate","Address a problem materially different from proposals 1 and 2","Use protected absence as the active perceptual and interpretive mechanism","Provide complete operations, authority, safeguards, rivals, falsifiers, and bounded evidence"],"conceptual_changes":[],"operational_changes":[],"evidence_changes":[],"claim_changes":[]}}