{ "schema_version": "2.0", "project_id": "HG-SYN-001", "status": "draft", "updated_on": "2026-07-23", "human_owner": "Accountable human investigator", "observation": { "statement": "In a synthetic demonstration dataset, units measured after condition X had a different mean value of outcome Y than units measured after the comparator condition.", "provenance": "Synthetic demonstration only; no human, animal, clinical, proprietary, or security-sensitive data.", "source_ids": [ "SRC-SYN-001" ], "uncertainties": [ "The synthetic pattern could reflect the generating process, selection, measurement, or random variation." ] }, "research_question": { "statement": "In the synthetic target system, what is the effect of assigning condition X rather than the comparator on outcome Y after one measurement interval?", "framework": "PICO", "question_type": "causal", "population_or_system": "Synthetic independent units meeting the declared eligibility rule", "intervention_or_exposure": "Assignment to synthetic condition X", "comparator": "Assignment to the synthetic comparator condition", "outcome": "Validated synthetic outcome Y", "timeframe": "One prespecified measurement interval" }, "hypotheses": [ { "hypothesis_id": "H1", "statement": "Condition X changes outcome Y through the proposed synthetic process M.", "mechanism": "Condition X changes intermediate M before outcome Y is measured.", "status": "candidate", "source_ids": [ "SRC-SYN-001" ], "assumptions": [ "The synthetic assignment and measurement procedures operate as declared." ], "boundary_conditions": [ "The candidate is limited to the declared synthetic units and one-interval horizon." ], "uncertainties": [ "The intermediate may be a marker rather than a mediator." ], "prediction_ids": [ "P1" ], "rival_hypothesis_ids": [ "H2" ] }, { "hypothesis_id": "H2", "statement": "The observed difference in outcome Y is produced by a synthetic measurement drift rather than process M.", "mechanism": "A time-linked offset changes the recorded outcome without changing the target construct.", "status": "candidate", "source_ids": [ "SRC-SYN-001" ], "assumptions": [ "The synthetic measurement offset is not removed by calibration." ], "boundary_conditions": [ "The candidate applies only to measurements from the affected synthetic acquisition period." ], "uncertainties": [ "The size and stability of any offset are unknown." ], "prediction_ids": [ "P2" ], "rival_hypothesis_ids": [ "H1" ] } ], "causal_estimands": [ { "estimand_id": "E1", "linked_hypothesis_ids": [ "H1" ], "population": "Synthetic eligible units", "intervention_or_exposure": "Assignment to condition X", "comparator": "Assignment to comparator", "outcome": "Outcome Y measured with method M-OUTCOME", "time_horizon": "One measurement interval", "population_summary": "Difference in population mean outcome Y", "intercurrent_event_strategy": "Not applicable in the synthetic demonstration", "identification_assumptions": [ "Random assignment, no interference, consistent treatment versions, and valid outcome measurement." ] } ], "predictions": [ { "prediction_id": "P1", "hypothesis_id": "H1", "statement": "Under matched acquisition conditions, assignment to X will precede a change in intermediate M and then outcome Y.", "observable": "Ordered changes in intermediate M and outcome Y", "conditions": "Synthetic units are randomly assigned and measured with calibrated instruments.", "expected_pattern": "Intermediate M changes before outcome Y in X but not comparator units.", "falsifier": "Outcome Y changes without the prespecified change in intermediate M under a successful manipulation and valid measurements.", "rival_hypothesis_ids": [ "H2" ], "measurement_ids": [ "M-INTERMEDIATE", "M-OUTCOME" ], "analysis_ids": [ "A1" ] }, { "prediction_id": "P2", "hypothesis_id": "H2", "statement": "A reference quantity measured in the same acquisition period will show a similar offset even though process M cannot affect it.", "observable": "Offset in the negative-control reference quantity", "conditions": "The same acquisition and calibration workflow is used.", "expected_pattern": "The reference quantity and outcome Y shift together by acquisition period.", "falsifier": "The reference quantity remains stable while outcome Y changes reproducibly across independently calibrated periods.", "rival_hypothesis_ids": [ "H1" ], "measurement_ids": [ "M-NEGATIVE-CONTROL", "M-OUTCOME" ], "analysis_ids": [ "A2" ] } ], "alternative_explanations": [ { "alternative_id": "ALT1", "statement": "Synthetic selection into the analyzed set creates the observed difference.", "linked_hypothesis_ids": [ "H1", "H2" ], "risk_types": [ "selection_bias" ], "discriminating_prediction_ids": [ "P1", "P2" ] } ], "null_hypotheses": [ { "null_id": "N1", "statement": "The population mean of outcome Y is equal under assignment to X and comparator at the prespecified horizon.", "linked_prediction_ids": [ "P1" ], "rejection_or_compatibility_rule": "Use the prespecified interval estimate and decision rule in analysis A1; do not equate failure to reject with equivalence." }, { "null_id": "N2", "statement": "The synthetic negative-control reference mean is equal across acquisition periods.", "linked_prediction_ids": [ "P2" ], "rejection_or_compatibility_rule": "Use analysis A2 as a bias diagnostic; failure to reject does not prove absence of drift when precision or reference sensitivity is insufficient." } ], "negative_controls": [ { "control_id": "NC1", "control_type": "negative_outcome", "rationale": "The reference quantity cannot be changed by process M but shares the synthetic acquisition workflow.", "expected_result": "No difference by assigned condition after accounting for acquisition period.", "failure_implication": "A non-null control result raises concern about acquisition, selection, or analysis bias.", "linked_prediction_ids": [ "P2" ] } ], "operationalizations": [ { "measurement_id": "M-INTERMEDIATE", "construct": "Synthetic intermediate M", "variable": "m_value", "role": "mediator", "operational_definition": "Calibrated continuous synthetic measurement at the prespecified early time.", "instrument_or_method": "Deterministic synthetic measurement procedure", "unit": "synthetic units", "timing": "Early prespecified time", "population_or_system": "Synthetic eligible units", "validity_evidence_source_ids": [ "SRC-SYN-001" ], "reliability_plan": "Repeat the deterministic check on an independent synthetic batch.", "missingness_plan": "Report missingness by assignment and use the prespecified sensitivity analysis.", "blinding_or_masking": "Synthetic analyst label is masked until validation completes.", "threshold_rationale": "No threshold; analyze the continuous measurement." }, { "measurement_id": "M-OUTCOME", "construct": "Synthetic outcome Y", "variable": "y_value", "role": "outcome", "operational_definition": "Calibrated continuous synthetic outcome at one interval.", "instrument_or_method": "Deterministic synthetic measurement procedure", "unit": "synthetic units", "timing": "One prespecified measurement interval", "population_or_system": "Synthetic eligible units", "validity_evidence_source_ids": [ "SRC-SYN-001" ], "reliability_plan": "Verify calibration and repeatability before unmasking assignment.", "missingness_plan": "Report reasons and analyze under the prespecified primary and sensitivity assumptions.", "blinding_or_masking": "Outcome acquisition and validation are masked to assignment.", "threshold_rationale": "No threshold; estimate the continuous mean contrast." }, { "measurement_id": "M-NEGATIVE-CONTROL", "construct": "Synthetic reference quantity", "variable": "reference_value", "role": "negative_control", "operational_definition": "Stable synthetic reference measured in the same acquisition workflow.", "instrument_or_method": "Deterministic synthetic reference procedure", "unit": "synthetic reference units", "timing": "Concurrent with outcome Y", "population_or_system": "Synthetic acquisition batches", "validity_evidence_source_ids": [ "SRC-SYN-001" ], "reliability_plan": "Check repeatability across synthetic batches.", "missingness_plan": "Treat a missing reference as an acquisition-quality failure.", "blinding_or_masking": "Reference processing is independent of assignment labels.", "threshold_rationale": "Use the prespecified calibration tolerance." } ], "analysis_plan": { "analyses": [ { "analysis_id": "A1", "prediction_ids": [ "P1" ], "estimand_ids": [ "E1" ], "analysis_population": "All randomized synthetic eligible units under the declared estimand strategy", "method": "Estimate the prespecified mean contrast with its uncertainty interval.", "effect_or_summary_measure": "Difference in population means", "uncertainty_method": "Prespecified interval estimate under the declared independent-unit assumptions", "missing_data_plan": "Primary declared assumption plus a bounded sensitivity analysis", "multiplicity_plan": "A1 is the sole confirmatory contrast; all additional contrasts are exploratory.", "sensitivity_analyses": [ "Repeat under the prespecified alternative missingness assumption." ], "decision_rule": "Interpret compatibility with predicted patterns and rivals; do not select a hypothesis automatically.", "exploratory_or_confirmatory": "confirmatory" }, { "analysis_id": "A2", "prediction_ids": [ "P2" ], "estimand_ids": [], "analysis_population": "All synthetic acquisition batches with valid reference measurements", "method": "Estimate reference shifts by acquisition period and assigned condition.", "effect_or_summary_measure": "Difference in reference means", "uncertainty_method": "Prespecified interval estimate", "missing_data_plan": "Report and exclude acquisition failures according to the locked quality rule.", "multiplicity_plan": "Negative-control analysis is interpreted as a bias diagnostic, not a candidate-selection score.", "sensitivity_analyses": [ "Repeat after excluding the prespecified failed-calibration batch category." ], "decision_rule": "A non-null reference shift triggers investigation and limits causal interpretation.", "exploratory_or_confirmatory": "confirmatory" } ], "harking_control": "Freeze this record before viewing the target synthetic outcomes and label later candidates or analyses exploratory.", "deviation_reporting": "Preserve the original plan and log each deviation with date, rationale, decision owner, result-awareness status, and interpretive impact." }, "evidence": { "ledger_path": "local-evidence-ledger.csv", "source_ids": [ "SRC-SYN-001" ], "search_boundary_id": "SEARCH-SYN-001", "evidence_limitations": [ "The bundled row is synthetic and demonstrates structure only." ] }, "risk_register": { "confounding": [ "Synthetic assignment failure or implementation differences could act as common causes." ], "selection_bias": [ "Post-assignment exclusion could make analyzed groups non-comparable." ], "collider_bias": [ "Restricting analysis to a jointly caused quality indicator could open bias." ], "reverse_causation": [ "Temporal ordering is built into the synthetic design but must be verified." ], "measurement_bias": [ "Acquisition-period drift could change recorded outcome Y." ], "other": [ "Random variation and model misspecification remain possible." ] }, "ethics_and_feasibility": { "human_subjects_gate": "not_applicable", "animal_research_gate": "not_applicable", "biosafety_gate": "not_applicable", "dual_use_gate": "not_applicable", "regulatory_gate": "not_applicable", "data_governance_gate": "approved", "feasibility_status": "feasible_for_planning", "required_reviews": [], "unresolved_blocks": [] }, "ai_use": { "used": false, "sensitive_or_unpublished_data_sent_externally": false, "local_policy_checked": true, "citation_verification_required": true, "human_accountable": true, "diversity_mitigation": "Generate independent rivals before any authorized AI-assisted expansion and retain human-origin candidates." } }