{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp06_four_proposal_generalization60_20260803","cell_id":"predictive_residual_processing__environmental_climate","arm":"COMPLETE_PROPOSAL_PORTFOLIO","candidate_id":"prp-command-conditioned-river-residual-watch-004","proposal_index":4,"version":0,"title":"Command-Conditioned Residual Watch for Managed River Releases","problem":"During a controlled reservoir release, downstream stage, temperature, dissolved oxygen, conductivity, and turbidity can change substantially because of gate and turbine operations. These expected self-caused responses dominate raw monitoring displays and fixed alarms, making it difficult to distinguish routine release effects from an actuator discrepancy, tributary contaminant pulse, bank failure, unexpected sediment mobilization, or other external disturbance occurring at the same time. Simply relaxing alarms during releases risks treating the operational period as an environmental blind spot.","actors":["Reservoir control-room operator","River environmental-monitoring lead","Downstream field technician","Hydrologic model owner","Environmental incident coordinator","Independent compliance-data custodian"],"observable_state":"For every authorized gate or turbine command, the system records the requested command, measured actuator response, timestamp, and provenance. At each downstream monitoring interval it retains the versioned command-conditioned prediction, predicted arrival window, raw sensor vector, signed residual, uncertainty, source-quality flags, propagation decision, reconstructed vector, heartbeat, audit status, and fallback state. Missing observations, failed actuators, model mismatches, and quiet residuals remain distinguishable states.","consequence":"Operators may spend attention repeatedly interpreting predictable release-driven fluctuations, while an unexplained environmental change is delayed, misattributed to the release, or hidden by a temporarily widened threshold. Conversely, an expected release response can repeatedly trigger raw alarms and normalize alarm dismissal.","affected_objective":"Distinguish unexplained downstream environmental changes from the predicted consequences of authorized reservoir operations while preserving complete safety, compliance, and scientific records and preventing the predictor from excusing consequential observations.","intervention":"Copy each outgoing gate or turbine command, pair it with measured actuation and declared upstream conditions, and pass it through a versioned forward model of the time-delayed downstream response at specified monitoring stations. Commit predicted stage, temperature, dissolved oxygen, conductivity, and turbidity envelopes before the corresponding observations arrive. Compare raw observations with those expectations, preserving signed residuals and timing discrepancies. Route precision- and consequence-weighted residuals to the environmental-response queue while keeping the predicted self-caused component implicit in that attention channel and reconstructible from the model version plus residual. Treat divergence between commanded and measured actuation as a residual rather than assuming the command occurred. Retain the complete raw stream independently, audit random and release-transition windows, and force full-signal review during checksum conflict, missing heartbeat, unmodeled inflow, excessive uncertainty, persistent structured residuals, or reconstruction failure. Dam-safety variables, legally required records, critical dissolved-oxygen or stage states, sensor faults, and public-warning conditions bypass cancellation. Residuals may inform a reviewed model recalibration or investigation but cannot directly change gates or emergency actions.","structural_mapping":[{"archetype_element":"Prediction target definition","domain_realization":"For each declared downstream station and release regime, the target is the next time-indexed vector of stage, temperature, dissolved oxygen, conductivity, and turbidity, including units, travel-time horizon, sampling interval, and variable-specific reconstruction tolerance."},{"archetype_element":"Predictive feedforward model","domain_realization":"A copy of the outgoing release command, measured actuator state, upstream inflow, and allowed environmental context enters a forward response model before downstream observations arrive."},{"archetype_element":"Expected and actual behavior","domain_realization":"The model commits a predicted downstream trajectory and uncertainty envelope; independently retained sensors supply the actual trajectory with calibration, quality, timestamp, and missingness metadata."},{"archetype_element":"Prediction comparator and error signal","domain_realization":"The comparator preserves observed-minus-predicted values, arrival-time errors, cross-station inconsistencies, command-actuation discrepancies, and structured missingness rather than producing a single nominal or alarm state."},{"archetype_element":"Efference-copy cancellation","domain_realization":"The predicted environmental consequence of the reservoir's own authorized action is removed from the scarce operator-attention stream, leaving the portion not explained by the command-conditioned response."},{"archetype_element":"Precision and consequence weighting","domain_realization":"Residual priority accounts for sensor uncertainty, actuator verification, travel-time uncertainty, persistence, cross-station coherence, variable-specific consequence, and the risk of coincident external disturbance."},{"archetype_element":"Residual reconstruction and provenance","domain_realization":"Every routed residual names the command record, actuator observation, forward-model checksum, station, timestamp, and context needed to recover the full observed value as prediction plus residual."},{"archetype_element":"Bounded update rule","domain_realization":"Validated residuals can revise response-model uncertainty or parameters only through an offline, versioned review that distinguishes external events, actuator error, sensor error, and changes caused by prior interventions."},{"archetype_element":"Independent raw audit","domain_realization":"A separate archive preserves all raw observations and samples complete random windows, release starts and stops, high-uncertainty periods, and coincident tributary events without using the cancellation model as ground truth."},{"archetype_element":"Fallback and protected bypass","domain_realization":"Model incompatibility, staleness, unverified actuation, novel hydrologic conditions, structured drift, missing heartbeat, or audit disagreement restores complete-signal attention. Safety, compliance, public-warning, and designated ecological-limit signals always travel in full."}],"mechanism_mapping":[{"mechanism_slug":"efference_copy_cancellation","role":"Uses a copy of the reservoir command to predict and subtract the operation's downstream sensory consequence so unexplained environmental change remains prominent.","counterfactual_removal":"Without the command copy and forward cancellation, the proposal becomes generic anomaly detection and cannot separate self-generated release responses from external disturbances."},{"mechanism_slug":"innovation_residual_filter","role":"Maintains a time-evolving downstream state and uncertainty estimate, applying observed-minus-predicted innovations while accounting for process, travel-time, and sensor uncertainty.","counterfactual_removal":"Without uncertainty-aware innovation processing, noisy observations and uncertain arrival timing can dominate the residual stream or produce unjustified confidence in cancellation."},{"mechanism_slug":"event_triggered_residual_reporting","role":"Sends validated unexplained departures to the operator-attention channel while authenticated heartbeats verify quiet intervals.","counterfactual_removal":"Without event-triggered routing, residuals remain one more full-volume display and do not relieve the attention competition created by predictable release effects."},{"mechanism_slug":"precision_weighted_error_gate","role":"Prioritizes residuals according to reliability, consequence, persistence, coherence, and attention cost while retaining suppressed residuals for audit.","counterfactual_removal":"Without the gate, large uncertain hydraulic errors can crowd out smaller reliable dissolved-oxygen, conductivity, or actuator discrepancies."},{"mechanism_slug":"forecast_backtesting","role":"Defines the release regimes, stations, variables, seasons, and travel-time horizons in which command-conditioned cancellation is permitted using held-out release records.","counterfactual_removal":"Without scoped backtesting, a response model may be used during hydrologic or operating conditions it has not demonstrated it can reconstruct."},{"mechanism_slug":"model_version_checksum_handshake","role":"Prevents a residual generated from one command-response model or station configuration from being interpreted against another.","counterfactual_removal":"Without compatibility checks, a well-formed residual can reconstruct the wrong downstream state after model, rating-curve, or station changes."},{"mechanism_slug":"model_drift_monitoring","role":"Tracks residual bias, autocorrelation, travel-time error, calibration, source conditions, and model age for changes in channel morphology, season, operations, or sensors.","counterfactual_removal":"Without drift monitoring, changing river dynamics can cause the cancellation model to become confidently wrong while continuing to suppress observations."},{"mechanism_slug":"shadow_raw_channel_sampling","role":"Compares reconstructed windows with independently retained complete observations across random periods and release transitions.","counterfactual_removal":"Without raw sampling, external signals consistently over-cancelled by the response model cannot be observed by the residual pipeline itself."},{"mechanism_slug":"raw_signal_fallback_switch","role":"Restores complete downstream observations when actuation, validity, reconstruction, coverage, safety, or compatibility checks fail.","counterfactual_removal":"Without fallback, the possibly invalid cancellation model remains between operators and the evidence needed to diagnose its failure."},{"mechanism_slug":"prediction_error_replay_buffer","role":"Stores residuals with command, actuator, upstream, station, and raw-window context for incident reconstruction, calibration, and regression testing.","counterfactual_removal":"Without contextual replay, later analysis cannot distinguish external events from timing, actuation, sensor, or model errors or test whether a revised model conceals prior failures."},{"mechanism_slug":"surprise_to_action_bridge","role":"Assigns validated unexplained changes to a named environmental or operational owner with full context, a defined investigation step, and acknowledgement.","counterfactual_removal":"Without an owned handoff, residuals can appear on a display without producing verification, sampling, or incident assessment."}],"causal_chain":["An authorized reservoir command is copied before execution and paired with independently measured actuator behavior.","A frozen forward model converts the verified operation into time- and station-specific predictions of downstream environmental response.","Downstream sensors capture the complete realized response with provenance and explicit missingness.","Signed innovations separate the predicted self-caused component from the portion unexplained by the command, while actuator discrepancies remain visible.","Uncertainty and consequence weighting routes reliable, coherent, or protected residuals to scarce operator attention.","The receiver reconstructs the full observation from the compatible prediction and residual before interpreting or investigating the departure.","A validated residual prompts a named investigation into external input, unexpected environmental response, actuator behavior, sensor condition, or model failure rather than being automatically attributed to the dam or the river.","Independent raw windows test whether cancellation erased coincident external events or consequential expected states.","Drift, incompatibility, excessive uncertainty, or audit failure restores complete-signal review; validated learning produces a separately approved model version."],"baseline":"A raw downstream monitoring dashboard with fixed variable thresholds and a separately displayed release schedule. During releases, operators visually interpret whether alarms appear operation-related and may use manually widened thresholds or annotations. Complete records remain available, but no command-conditioned prediction is subtracted, residuals are not reconstructive teaching signals, and audit sampling does not test over-cancellation.","nearest_rivals":["A generic anomaly detector can flag departures from historical normal but may treat an authorized release as anomalous because it does not receive an efference copy of the operating command.","Static release-mode thresholds widen or shift alarm bounds during operations but cannot represent time-delayed, multistation responses or expose command-actuation discrepancies and cumulative model error.","A hydrodynamic forecast dashboard displays predicted and observed trajectories together but does not make the unexplained residual the governed attention channel or require independent tests of what expected trajectories suppress.","Offline source-apportionment analysis may investigate an incident after collection but does not continuously cancel verified self-generated effects, synchronize predictor versions, or trigger immediate full-signal fallback.","A fixed before-and-after comparison attributes changes around a release to the operation but cannot distinguish coincident external disturbances, travel-time variation, or unexecuted commands."],"remaining_contrastive_claim":"The candidate is a governed environmental analogue of efference-copy cancellation: verified operating commands generate a reconstructible prediction of their downstream sensory consequences, and only the uncertainty-weighted unexplained remainder competes for routine response attention, while raw independent records and protected bypasses prevent expected operational effects from becoming an excuse for environmental harm.","authority_safety":{"decision_authority":"The river environmental-monitoring lead may authorize a retrospective and shadow evaluation of residual routing. The reservoir operator retains gate and turbine authority under existing procedures, the incident coordinator controls environmental response, and the designated safety authority controls emergency warnings and dam-safety actions.","authorized_first_step":"Replay one bounded historical release using frozen command, actuator, upstream, and downstream records, then observe one scheduled release in shadow mode while the existing raw dashboard, thresholds, staffing, and operating procedures remain authoritative.","excluded_actions":["Automatically open, close, or otherwise alter gates, turbines, spillways, or release schedules","Suppress, delete, downsample, or replace mandatory raw safety, compliance, or scientific records","Cancel a critical dissolved-oxygen, stage, equipment, or public-warning signal as an expected release effect","Use a quiet residual channel as proof of compliance, environmental safety, or absence of an external event","Reclassify an unverified command as executed","Tune cancellation strength or thresholds to reduce alarm counts","Update the response model during an evaluated release without a new version and review","Delay an established emergency or field-verification procedure pending model interpretation"],"halt_rollback":"Disable cancellation and restore the unchanged raw dashboard for the affected stations and variables upon any protected-signal suppression, command-actuator mismatch not surfaced, checksum conflict, missing heartbeat, failed reconstruction, unmodeled hydrologic condition, repeated fallback oscillation, or audit evidence of over-cancellation. Preserve raw data and residual logs and reinstate the last approved model and threshold versions before any retest."},"negative_tests":{"strongest_counterevidence":"The raw-dashboard baseline allows operators to distinguish release effects from consequential external or equipment events with equal or better timeliness and context at lower total modeling, auditing, and review cost, while command-conditioned cancellation hides coincident events or creates misleading causal attribution.","problem_falsifier":"Managed operations do not account for a substantial portion of the relevant downstream variation during review periods, raw alarms do not compete materially for attention, or command-to-observation responses are too nonstationary or weakly identified to support a defensible prediction and reconstruction envelope.","intervention_falsifier":"Held-out or scripted coincident external pulses are cancelled as expected release effects; commanded and measured actuation discrepancies remain hidden; audit-window reconstruction exceeds declared tolerances; protected states fail to bypass; missing data appears as agreement; travel-time or residual drift fails to trigger decompression; or total model, audit, and response burden exceeds the raw baseline at required completeness.","risks":["An external contaminant or sediment pulse coincident with a release could be over-cancelled.","A recorded command may differ from physical gate or turbine behavior, corrupting the predicted consequence.","Incorrect travel time can convert a valid prediction into misleading positive and negative residual pairs.","Seasonal flow, channel morphology, tributary inflow, or reservoir stratification can invalidate the response model.","Predictable release effects can still be environmentally consequential and must not disappear from compliance or ecological-limit review.","Operators or model owners could tune the predictor to explain away known exceedances or operational errors.","Sensor faults or communication loss could resemble a quiet residual stream if heartbeat handling fails.","Updating from observations affected by response actions can teach the model its own intervention rather than river behavior.","Attention focused on residuals may remove baseline context needed to interpret ecological duration or cumulative exposure.","A shared upstream-data error can affect both prediction and residual attribution." ]},"next_evidence_step":"Preregister a bounded replay using one historical controlled release with synchronized command, actuator, upstream, and downstream records. Freeze the model version, station set, prediction horizons, uncertainty assumptions, variable-specific reconstruction tolerances, protected bypasses, and fallback rules; reserve part of the release record as an untouched evaluation interval. Add scripted tests for a command-actuator mismatch, sensor dropout, checksum conflict, travel-time shift, persistent residual bias, and superposed conductivity, turbidity, and dissolved-oxygen departures occurring during the predicted release response. Compare the residual attention stream with the unchanged raw-threshold baseline and inspect random plus release-transition raw windows. Record reconstruction error, protected-signal routing, unexplained-event routing, false attribution, audit disagreement, fallback behavior, operator acknowledgements, and total processing and review burden. If those checks pass, run one scheduled release in non-operational shadow mode; neither step establishes environmental safety or operational benefit.","prior_art_status":"UNSEARCHED","diversity_from_prior_proposals":"Proposal 1 addressed constrained transmission from passive peatland-rewetting sensors; its model suppressed predictable natural measurements to conserve station energy and communications capacity. This proposal does not primarily compress a remote link: it copies deliberate reservoir commands and cancels their predicted environmental consequences to distinguish self-generated response from coincident external change in an operator-attention channel. Proposal 2 addressed retrospective climate-model evaluation, propagating observation-model residuals through spatial and process layers to guide scientific model revision. This proposal instead operates around a live controlled intervention, uses command-conditioned temporal response and actuator verification rather than a scientific hierarchy, and ends in environmental or operational investigation rather than a climate-model acceptance decision. Proposal 3 addressed review of industrial emissions-renewal documents using a regulator-owned expected dossier and semantic or numerical filing residuals. This proposal processes physical river observations and verified operating commands, has no documentary approval baseline, and cannot approve or condition a regulated filing. It is independently adoptable by a reservoir and river-monitoring operation without changing remote peatland telemetry, climate-model development, or emissions-dossier review.","revision_record":{"parent_version":null,"progress_targets_addressed":["Fourth independently adoptable candidate","Material distinction from proposals 1, 2, and 3","Efference-copy causal path","Complete operational authority and safety controls","Explicit falsifiers and bounded evidence"],"conceptual_changes":["Instantiated predictive residual processing as separation of self-generated environmental consequences from external river disturbances","Made verified command-to-response prediction the source of expected input rather than passive history, a climate simulation, or a regulatory dossier"],"operational_changes":["Defined command and actuator provenance, time-delayed downstream prediction, protected full-signal bypasses, independent raw archives, scope-specific decompression, and non-automatic investigation handoffs"],"evidence_changes":["Specified a historical replay and one shadow release with held-out data, coincident-event tests, actuation mismatch, drift, missingness, version, reconstruction, and fallback checks"],"claim_changes":["Made no novelty, prevalence, demand, safety, compliance, or effect-size claim; limited the proposed evidence to feasibility and failure-handling tests"]}}