Forge Intelligence

Handbook

LCDL overlay: Experimentation and Empirical Learning Intelligence

Design and test-contract bundle; no runtime implementation or operational certification is claimed. The manifest, schemas, validators, test suite, and task identities are complete enough for repository-native…

Updated

Runtime boundary

Design and test-contract bundle; no runtime implementation or operational certification is claimed. The manifest, schemas, validators, test suite, and task identities are complete enough for repository-native implementation; no placeholder task code is included.

Capability contracts

Capability ID Candidate task ID Required output artifacts Release validators
capability.general.experimentation.formulate_testable_hypotheses intelligence_experimentation_formulate_testable_hypotheses_v1 artifact.general.experimentation.hypothesis_registry validator.general.experimentation.01_hypothesis_design_confounding_control_and_causal_identification_audit, validator.general.experimentation.02_randomization_sample_calculation_model_and_result_recomputation, validator.general.experimentation.03_schema_missingness_transformation_protocol_deviation_and_lineage_checks
capability.general.experimentation.define_variables_and_measures intelligence_experimentation_define_variables_and_measures_v1 artifact.general.experimentation.variable_and_measurement_dictionary validator.general.experimentation.01_hypothesis_design_confounding_control_and_causal_identification_audit, validator.general.experimentation.02_randomization_sample_calculation_model_and_result_recomputation, validator.general.experimentation.03_schema_missingness_transformation_protocol_deviation_and_lineage_checks
capability.general.experimentation.select_experimental_design intelligence_experimentation_select_experimental_design_v1 artifact.general.experimentation.experimental_design_matrix validator.general.experimentation.01_hypothesis_design_confounding_control_and_causal_identification_audit, validator.general.experimentation.02_randomization_sample_calculation_model_and_result_recomputation, validator.general.experimentation.03_schema_missingness_transformation_protocol_deviation_and_lineage_checks
capability.general.experimentation.control_confounding_and_bias intelligence_experimentation_control_confounding_and_bias_v1 artifact.general.experimentation.confounding_and_bias_control_plan validator.general.experimentation.02_randomization_sample_calculation_model_and_result_recomputation, validator.general.experimentation.03_schema_missingness_transformation_protocol_deviation_and_lineage_checks
capability.general.experimentation.plan_sampling_and_stopping intelligence_experimentation_plan_sampling_and_stopping_v1 artifact.general.experimentation.sampling_and_stop_plan validator.general.experimentation.02_randomization_sample_calculation_model_and_result_recomputation
capability.general.experimentation.author_protocol_and_analysis_plan intelligence_experimentation_author_protocol_and_analysis_plan_v1 artifact.general.experimentation.protocol_and_analysis_plan validator.general.experimentation.03_schema_missingness_transformation_protocol_deviation_and_lineage_checks, validator.general.experimentation.02_randomization_sample_calculation_model_and_result_recomputation, validator.general.experimentation.04_consent_safety_privacy_vulnerable_subject_and_regulated_research_gate
capability.general.experimentation.execute_and_log_experiment intelligence_experimentation_execute_and_log_experiment_v1 artifact.general.experimentation.experiment_run_log, artifact.general.experimentation.data_and_transformation_manifest validator.general.experimentation.03_schema_missingness_transformation_protocol_deviation_and_lineage_checks, validator.general.experimentation.05_protocol_code_data_environment_seed_and_replication_readiness_audit
capability.general.experimentation.analyze_results_and_uncertainty intelligence_experimentation_analyze_results_and_uncertainty_v1 artifact.general.experimentation.result_and_uncertainty_report validator.general.experimentation.02_randomization_sample_calculation_model_and_result_recomputation
capability.general.experimentation.assess_validity_replicate_and_iterate intelligence_experimentation_assess_validity_replicate_and_iterate_v1 artifact.general.experimentation.validity_replication_and_next_experiment_plan validator.general.experimentation.01_hypothesis_design_confounding_control_and_causal_identification_audit, validator.general.experimentation.02_randomization_sample_calculation_model_and_result_recomputation, validator.general.experimentation.05_protocol_code_data_environment_seed_and_replication_readiness_audit

Each capability contract in the manifest defines inputs, outputs, preconditions, postconditions, method-selection rules, invariant references, failure states, risk gate, metrics, and the existing forge-lcdl tasks/operators to compose.

Artifact schemas

Artifact type Schema reference Required payload fields Sensitivity
artifact.general.experimentation.hypothesis_registry ../schemas/artifacts/experimentation.schema.json#/$defs/hypothesis_registry hypothesis_id, statement, rival_hypotheses, predicted_observations, falsifiers, assumptions, status internal
artifact.general.experimentation.variable_and_measurement_dictionary ../schemas/artifacts/experimentation.schema.json#/$defs/variable_and_measurement_dictionary entries, stable_identifiers, definitions_or_types, units_or_formats, provenance, version, validation_status internal
artifact.general.experimentation.experimental_design_matrix ../schemas/artifacts/experimentation.schema.json#/$defs/experimental_design_matrix purpose, scope, entities_or_nodes, relations_or_edges, assumptions, boundary, validation_evidence internal
artifact.general.experimentation.confounding_and_bias_control_plan ../schemas/artifacts/experimentation.schema.json#/$defs/confounding_and_bias_control_plan objective, scope, steps, dependencies, constraints, owners, timing, contingencies, review_triggers internal
artifact.general.experimentation.sampling_and_stop_plan ../schemas/artifacts/experimentation.schema.json#/$defs/sampling_and_stop_plan objective, scope, steps, dependencies, constraints, owners, timing, contingencies, review_triggers internal
artifact.general.experimentation.protocol_and_analysis_plan ../schemas/artifacts/experimentation.schema.json#/$defs/protocol_and_analysis_plan objective, scope, steps, dependencies, constraints, owners, timing, contingencies, review_triggers internal
artifact.general.experimentation.experiment_run_log ../schemas/artifacts/experimentation.schema.json#/$defs/experiment_run_log entries, entry_ids, owners, status, source_refs, updated_at internal
artifact.general.experimentation.data_and_transformation_manifest ../schemas/artifacts/experimentation.schema.json#/$defs/data_and_transformation_manifest entries, stable_identifiers, definitions_or_types, units_or_formats, provenance, version, validation_status internal
artifact.general.experimentation.result_and_uncertainty_report ../schemas/artifacts/experimentation.schema.json#/$defs/result_and_uncertainty_report question_or_standard, criteria, evidence, findings, uncertainty, limitations, status, recommended_actions internal
artifact.general.experimentation.validity_replication_and_next_experiment_plan ../schemas/artifacts/experimentation.schema.json#/$defs/validity_replication_and_next_experiment_plan objective, scope, steps, dependencies, constraints, owners, timing, contingencies, review_triggers internal

Validators

Validator ID Kind Automation Severity Failure action
validator.general.experimentation.01_hypothesis_design_confounding_control_and_causal_identification_audit design_review hybrid major repair
validator.general.experimentation.02_randomization_sample_calculation_model_and_result_recomputation deterministic_statistical deterministic major repair
validator.general.experimentation.03_schema_missingness_transformation_protocol_deviation_and_lineage_checks data_integrity hybrid major repair
validator.general.experimentation.04_consent_safety_privacy_vulnerable_subject_and_regulated_research_gate ethics_policy hybrid minor conditional_pass
validator.general.experimentation.05_protocol_code_data_environment_seed_and_replication_readiness_audit reproducibility hybrid minor conditional_pass

Validator checks return itemized evidence and support pass, conditional_pass, fail, inconclusive, or a governed repair/escalation path. Generation does not self-certify.

Existing execution substrate

Tasks to reuse: contradiction_scan, decompose_problem, extract_schema_from_text, formula_eval_safe, llm_boolean_gate, llm_enum_route, llm_multi_label, plan_decision_pack, word_problem_to_calc_plan
Operators to reuse: branch, fallback_chain, for_each, optional_step, seq, try_catch, until_ok

Cross-intelligence handoff

  • artifact_id, type, version, producer capability, provenance, scope, assumptions, uncertainty, sensitivity, validation status, and next action
  • No free-text handoff may replace a required typed artifact when a downstream validator depends on structure.

Information-loss checks: - all hard constraints preserved - all blocker findings preserved - source scope and date preserved - assumptions and dissent preserved - units, time zones, identifiers, and uncertainty preserved

Evaluation and release

Suite: ../evaluation_suites/experimentation.yaml

  • 100% schema validity for released artifacts
  • 100% material traceability
  • no unresolved blocker validator
  • no missed critical escalation in the approved benchmark
  • documented conditional-pass limits and human gates

Promotion to implementation maturity I1

  • Create repository-native input and output contracts for every candidate task ID.
  • Implement deterministic portions of validators and declare independent human or model review for semantic portions.
  • Connect a provenance-preserving source resolver and enforce freshness and jurisdiction rules.
  • Pass the per-pack suite and dependency regression suites at release thresholds.
  • Complete security, privacy, accessibility, and domain-owner review appropriate to risk.
  • Publish LMETA/LCDL exact-overlay links and CI parity checks.

Exact-overlay identity

The LMETA and LCDL pages share the concept ID, semantic version, capability IDs, and invariant IDs. tools/validate_pack.py checks this parity.

Paired overlay

Runtime maturity

  • Capabilities: 9
  • Technical floor→ceiling: I1_runtime → I2_integrated
  • Autonomy floor→ceiling: A0_assisted → A0_assisted
  • I1_runtime count: 8
  • Evidence class: contract: 8, semantic: 1
  • Full hierarchy: MATURITY-HIERARCHY.md

Evaluation suite status

  • CI status: pass
  • Suite: evaluation_suites/experimentation.yaml
  • Cases defined: 7
  • Capabilities under test: 9