{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp06_four_proposal_generalization60_20260803","cell_id":"predictive_residual_processing__sociology_anthropology","arm":"COMPLETE_PROPOSAL_PORTFOLIO","candidate_id":"prp_delta_relationship_mapping_003","proposal_index":3,"version":0,"title":"Delta-Elicited Relationship Maps for Longitudinal Community Fieldwork","problem":"A longitudinal community study repeatedly maps participants' kinship, caregiving, resource-sharing, and mutual-aid relationships. Re-eliciting every authorized relationship at every wave consumes interview time and repeatedly asks participants to disclose stable sensitive information, while simply carrying the previous roster forward can miss new, ended, reclassified, or uncertain ties. A basic change-only question is also unsafe because silence can mean unchanged, forgotten, withheld, misunderstood, or not reviewed.","actors":["Consented study participants reporting their own relationships","Community interviewers","Longitudinal network and ethnographic analysts","Research data steward","Study principal investigator","Community advisory group","Research ethics authority"],"observable_state":"For each participant and wave, the study can compare the last participant-confirmed relationship snapshot with the next complete authorized roster. Observable quantities include unchanged and changed fields, additions and removals, interview time, missing responses, corrections, reconstruction disagreement, snapshot age, version mismatch, participant refusals, and differences across tie classes or participant groups.","consequence":"Full re-enumeration may impose unnecessary interview and disclosure burden, but ungoverned carry-forward or informal change-only elicitation can create stale relationship maps, convert missingness into false continuity, and obscure socially consequential changes in care or obligation.","affected_objective":"Maintain longitudinal relationship maps at the fidelity required for sociological and anthropological analysis while reducing repeated elicitation and preserving participant control, explicit missingness, contextual interpretation, and access to full re-enumeration.","intervention":"For each participating individual, generate a private, versioned expected-state snapshot containing only relationships that person previously confirmed and remains authorized to review. At the next wave, present the snapshot in human-readable form and ask the participant to encode structured residuals: add a tie, end a tie, change its type or direction, revise activity or confidence, correct the prior record, mark unknown, decline, or request full re-enumeration. The interview system reconstructs the proposed current map as the prior snapshot plus confirmed residuals and asks the participant to approve the reconstruction. Residuals carry snapshot version, field provenance, uncertainty, and consent scope. High uncertainty, stale snapshots, extensive change, participant discomfort, household transition, safeguarding concerns, or checksum mismatch trigger a complete roster interview. A random sample and selected high-risk strata receive full re-enumeration regardless of apparent stability, and all participants receive periodic full refreshes. Operational records update only from confirmed responses; predictive assumptions about which fields remain stable update separately and only after reviewed audit evidence.","structural_mapping":[{"archetype_element":"Prediction target and observation boundary","domain_realization":"Predict the participant-confirmed state of a declared set of kinship, care, exchange, and mutual-aid ties at the next interview wave; exclude unconsented persons, inferred hidden ties, motives, and relationship value judgments."},{"archetype_element":"Generative model state","domain_realization":"The expected state combines the participant's last confirmed snapshot with explicit assumptions about field persistence, snapshot age, and uncertainty, all under a versioned model owned by the study."},{"archetype_element":"Expected behavior","domain_realization":"The participant-specific prior snapshot supplies the expected roster, tie categories, directions, activity states, and unresolved fields for the new wave."},{"archetype_element":"Actual behavior","domain_realization":"The participant supplies explicit confirmations, edits, additions, removals, uncertainty, refusals, and full-reenumeration requests with timestamp and interviewer provenance."},{"archetype_element":"Prediction comparator","domain_realization":"The comparator represents relationship additions, removals, category substitutions, direction changes, activity changes, corrected prior errors, unknown states, refusals, and unreviewed fields."},{"archetype_element":"Prediction-error signal","domain_realization":"The structured edits are the residual payload; unchanged content remains implicit only after explicit review status permits reconstruction."},{"archetype_element":"Precision and consequence weighting","domain_realization":"Routing and fallback depend on snapshot age, response confidence, interviewer quality, tie sensitivity, analytic consequence, subgroup coverage, and the cost of carrying an incorrect relationship forward."},{"archetype_element":"Residual propagation and reconstruction","domain_realization":"The data system applies confirmed residuals to the matching snapshot and presents the reconstructed current map for participant confirmation before it becomes the operational record."},{"archetype_element":"Model synchronization and provenance","domain_realization":"Participant-facing form and research database must reference the same snapshot checksum, consent scope, category definitions, and model version; mismatches block delta application."},{"archetype_element":"Freshness and validity window","domain_realization":"Every snapshot expires after a declared interval or specified life transition, regardless of whether residuals have been reported."},{"archetype_element":"Bounded update rule","domain_realization":"Confirmed edits update that participant's relationship state immediately, while general persistence assumptions and fallback thresholds change only through scheduled, attributable audit review."},{"archetype_element":"Residual error budget","domain_realization":"The study limits incorrect carry-forwards, unreconstructable fields, consequential tie-class errors, and cumulative uncertainty rather than optimizing only the number of questions omitted."},{"archetype_element":"Raw-signal audit and resynchronization","domain_realization":"Random, risk-stratified, and periodic full roster interviews provide independent comparisons and complete snapshots that reset accumulated error."},{"archetype_element":"Fallback and safety bypass","domain_realization":"Participant request, discomfort, consent ambiguity, safeguarding concern, extensive change, staleness, or version failure suspends delta elicitation and opens a full or specially governed interview path."},{"archetype_element":"Attention and bandwidth budget","domain_realization":"The scarce resources are participant interview time, repeated sensitive disclosure, interviewer capacity, and data-processing effort; model, audit, and fallback costs count against any saving."},{"archetype_element":"Silence semantics","domain_realization":"Each field distinguishes confirmed unchanged, changed, unknown, declined, not reviewed, and missing interview; absence of a residual never automatically means continuity."}],"mechanism_mapping":[{"mechanism_slug":"delta_or_differential_encoding","role":"Represents the new participant-confirmed relationship map as structured edits against the last authorized snapshot rather than retransmitting every stable tie.","counterfactual_removal":"Without reconstructive deltas, the protocol becomes ordinary full roster elicitation or an informal change question with no guaranteed current-state reconstruction."},{"mechanism_slug":"predictive_codec","role":"Maintains compatible participant-facing and database snapshots so confirmed residuals reconstruct the full current map against a known baseline.","counterfactual_removal":"Without matched encoding and decoding state, a valid relationship edit could be applied to the wrong roster and silently corrupt the map."},{"mechanism_slug":"model_version_checksum_handshake","role":"Checks snapshot identity, consent scope, category definitions, and active version before any residual is interpreted or stored.","counterfactual_removal":"Without the handshake, late forms, category revisions, or database drift could produce plausible but false reconstructions."},{"mechanism_slug":"event_triggered_residual_reporting","role":"Elicits and transmits changed, uncertain, corrected, or newly sensitive relationship fields while explicit review statuses verify apparent stability.","counterfactual_removal":"Without residual reporting, the expected snapshot would not channel current participant evidence into the longitudinal record."},{"mechanism_slug":"precision_weighted_error_gate","role":"Combines uncertainty, snapshot age, source quality, tie sensitivity, analytic consequence, and review cost to decide whether a delta is sufficient or fuller elicitation is required.","counterfactual_removal":"Without consequence-aware weighting, the system might spend effort on inconsequential wording changes while carrying forward an uncertain caregiving or dependency relationship."},{"mechanism_slug":"confidence_threshold_table","role":"Versions the rules mapping snapshot age, residual mass, uncertainty, tie class, and participant response into accept, clarify, audit, or fully re-enumerate actions.","counterfactual_removal":"Without explicit thresholds, interviewers could apply inconsistent fallback standards or tune them merely to shorten interviews."},{"mechanism_slug":"periodic_full_state_resynchronization","role":"Requires complete roster refreshes on a fixed maximum interval and after specified transitions so delta errors cannot accumulate indefinitely.","counterfactual_removal":"Without periodic full snapshots, forgotten additions, dropped residuals, and misclassified ties could persist across every later wave."},{"mechanism_slug":"shadow_raw_channel_sampling","role":"Uses random and risk-stratified full roster interviews to compare reconstructed maps with independently re-elicited current reports.","counterfactual_removal":"Without independent full elicitation, the study cannot measure relationships that the previous snapshot or residual prompt caused participants to overlook."},{"mechanism_slug":"model_drift_monitoring","role":"Tracks rising residual density, corrections, unknown states, subgroup-specific disagreement, snapshot staleness, and audit failures across waves.","counterfactual_removal":"Without drift monitoring, a period of rapid network reorganization could remain inside a protocol calibrated for stability."},{"mechanism_slug":"raw_signal_fallback_switch","role":"Reverts to complete roster elicitation when validity, safety, consent, comfort, reconstruction, or synchronization conditions fail.","counterfactual_removal":"Without an accessible full path, participants and interviewers would be forced to express major or sensitive changes through an invalid prior roster."}],"causal_chain":["A participant's last authorized, confirmed relationship map becomes a scoped and versioned prediction of the next-wave state.","The next interview exposes that exact snapshot privately and distinguishes reviewed fields from missing or unreviewed ones.","The participant supplies signed, structured changes and uncertainty rather than restating every stable relationship.","A version check ensures each residual is applied only to the snapshot and consent scope that generated it.","The system reconstructs the proposed full current map as prior state plus residuals and returns it to the participant for confirmation.","Precision, consequence, staleness, and change-density rules determine whether the reconstruction is accepted, clarified, audited, or replaced by full re-enumeration.","Confirmed state changes update the longitudinal map, while retained residuals reveal where persistence assumptions failed.","Random and periodic full roster interviews measure omitted change, prompting bias, reconstruction error, and subgroup-specific failure.","Drift or an exhausted error budget suspends delta mode for the affected scope and restores complete elicitation.","Reviewed audit evidence may revise persistence assumptions and thresholds without rewriting participants' confirmed relationship histories."],"baseline":"At every wave, interviewers ask participants to enumerate the complete authorized relationship roster and describe each tie again. Alternatively, some studies may carry prior values forward and ask a general change question. Full re-enumeration preserves a fresh explicit record but can repeat stable questions; informal carry-forward reduces questions but lacks reconstructive versioning, explicit missingness, independent raw audits, bounded error, and reliable fallback.","nearest_rivals":["Full roster re-enumeration: obtains an explicit current report at every wave but does not use a predictive residual channel to reduce repeated elicitation.","Simple panel prefill or confirmation forms: display prior answers but may treat unchecked or silent fields as unchanged without model versions, reconstruction tests, error budgets, or independent full refreshes.","Event-history calendars: help participants recall dated transitions but do not reconstruct an entire current relationship map from synchronized prior state plus residuals.","Network diaries: collect relationship events as they happen, but missing diary entries can be mistaken for stability and usually lack periodic full-state reconciliation.","Administrative-record change feeds: transmit recorded additions and removals efficiently but do not preserve participant authority, subjective relationship categories, uncertainty, or refusal states.","Adaptive questionnaires: select follow-up questions from prior answers, but need not encode the new record as a reconstructive residual or audit what the adaptive path suppressed."],"remaining_contrastive_claim":"The design's remaining contrastive claim is structural: a participant-authorized prior relationship snapshot functions as a versioned prediction; explicit edits are the reconstructive message; confirmed residuals update the state; and checksum matching, explicit silence states, independent full-roster audits, bounded drift, participant-controlled decompression, and periodic resynchronization govern when unchanged content may remain implicit.","authority_safety":{"decision_authority":"Each participant retains authority to disclose, correct, decline, or request full re-enumeration for relationships they are authorized to discuss. The principal investigator may approve only the research workflow within the ethics protocol; the data steward controls snapshot access and linkage; the ethics authority governs collection and consent changes; and the community advisory group may advise on categories and safeguards but cannot inspect individual records without separate authorization.","authorized_first_step":"Perform an offline shadow reconstruction using deidentified consecutive-wave roster data already authorized for analysis. Treat the earlier full roster as the predicted snapshot and encode the later full roster as residuals without contacting participants or changing the record of record.","excluded_actions":["Showing a participant relationships reported solely by another person","Adding model-inferred, suspected, or unconsented relationships to the participant-facing snapshot","Treating no response, refusal, uncertainty, or an incomplete interview as confirmation of continuity","Using relationship residuals to determine benefits, policing, immigration action, housing access, or participant eligibility","Ranking participants or relationships by social worth, normality, credibility, or conformity","Discarding complete authorized wave data after producing residuals","Automatically changing consent scope or disclosure permissions","Optimizing thresholds solely for shorter interviews or smaller datasets","Releasing identifiable high-residual cases without existing authorization","Forcing delta elicitation when a participant requests a fresh interview or blank form"],"halt_rollback":"Halt delta processing for the affected participant or scope if snapshot identity or consent cannot be verified, a participant-facing record contains another person's unauthorized disclosure, reconstruction fails, missingness is converted to continuity, audit errors exceed the preregistered budget, errors cluster by subgroup or tie class, residual density indicates instability, or model-and-audit costs erase the intended reduction in elicitation. Quarantine affected derived records, retain the authorized source data and audit trail, restore full roster elicitation or full-record analysis, notify the data steward and principal investigator, and obtain ethics review before reuse."},"negative_tests":{"strongest_counterevidence":"The strongest counterevidence would be that showing prior relationships anchors recall and interpretation: participants confirm visible ties, overlook absent ones, or preserve obsolete categories, causing delta-elicited maps to disagree systematically with independently elicited full rosters despite correct technical reconstruction.","problem_falsifier":"The inferred problem is falsified if full roster interviews are not meaningfully burdensome or repetitive, if relationship states change too extensively for a prior snapshot to compress the next wave, or if the required analytic and ethical fidelity demands independent full re-enumeration every time.","intervention_falsifier":"The intervention is falsified if total interview, model, synchronization, and audit effort is not lower than the baseline at the required fidelity, or if it increases omitted ties, stale carry-forwards, subgroup-specific errors, participant discomfort, unauthorized disclosure, reconstruction disagreement, or fallback burden.","risks":["Prior rosters may anchor recall and suppress newly relevant relationships.","Displaying an obsolete or sensitive relationship may distress or endanger a participant.","Relationship categories may impose administrative definitions on locally meaningful ties.","A participant-facing snapshot could reveal data supplied by someone else if provenance controls fail.","Rapid network changes may make residual elicitation longer or less reliable than a fresh roster.","Interviewers may treat a short residual list as evidence that the network is stable.","Random audits may miss rare but consequential relationship changes.","Risk-stratified audits may reproduce the model's existing assumptions about where change occurs.","High-residual participants may become identifiable or stigmatized in analytic outputs.","Snapshot and audit maintenance may consume more capacity than repeated full interviews.","A confirmed reconstruction can still omit relationships that the displayed prior roster failed to cue.","Pressure to reduce interview time may lead staff to discourage full re-enumeration requests."]},"next_evidence_step":"Pre-register an offline shadow test using up to 80 deidentified participant records with two consecutive complete roster waves already authorized for analysis. Lock the earlier roster as the versioned prediction, derive structured residuals from the later roster, reconstruct the later state, and compare it with the source record. Measure exact and field-level reconstruction disagreement, residual density, unknown and missing-state handling, consequential tie-class errors, subgroup-specific discrepancies, simulated fallback frequency, snapshot-age sensitivity, and total encoding plus audit effort. Include a random full-record audit and a deliberately corrupted-version test. Treat this step as a technical and structural feasibility test only; it does not test participant anchoring, comfort, or actual interview-time effects.","prior_art_status":"UNSEARCHED","diversity_from_prior_proposals":"Proposal 1 addresses delayed recognition of changing benefit-intake practices by routing model-relative deviations from full ethnographic encounter notes to analysts. Proposal 2 addresses causal ambiguity created by researchers' own interventions by predicting and cancelling expected self-generated consequences. Proposal 3 instead addresses repeated longitudinal elicitation of participant relationship maps: the participant is the authorized source of current state, the prior confirmed roster is the shared reconstructive baseline, and additions, removals, corrections, uncertainty, and refusals are the transmitted deltas. Its objective is current-state acquisition with lower repeated disclosure, not post-collection attention triage or attribution of researcher effects. It can be adopted without narrative fieldnote review or participant-facing research interventions, and neither earlier proposal supplies its participant-controlled reconstruction, roster synchronization, or periodic re-enumeration path.","revision_record":{"parent_version":null,"progress_targets_addressed":["Specified a third independently adoptable sociology-and-anthropology problem","Preserved prediction, residual encoding, reconstruction, update, synchronization, audit, drift, and fallback structure","Defined operational authority, consent safeguards, falsifiers, rivals, rollback, and bounded evidence","Explained material diversity from sealed proposals 1 and 2"],"conceptual_changes":["Instantiated the archetype as longitudinal relationship-state acquisition rather than analytic triage or reflexive causal attribution.","Made the participant-confirmed prior roster the prediction and participant-supplied relationship edits the reconstructive residual.","Separated participant-confirmed state updates from slower learning about field persistence and elicitation thresholds."],"operational_changes":["Specified private snapshots, checksums, explicit missingness states, participant confirmation, full-roster audits, expiration, and participant-controlled fallback.","Restricted the first step to offline shadow reconstruction using already-authorized consecutive-wave data.","Prohibited inferred ties, cross-participant disclosure, operational eligibility decisions, and forced delta elicitation."],"evidence_changes":["Defined an up-to-80-record reconstruction and error-budget test with random audits and corrupted-version testing.","Made anchoring, comfort, and actual interview burden explicit unresolved questions outside the first offline test."],"claim_changes":["Made no claim of novelty, prevalence, demand, or effect size.","Limited the proposal to a falsifiable participant-controlled residual elicitation architecture."]}}