{ "schema_version": "1.0.0", "package_id": "W08", "title": "From Observation to Identification", "language": "en", "phase": "Explain", "essential_question": "Can the change be attributed to the intervention?", "research_cutoff": "2026-08-24", "status": "structurally_complete_text_first_authored_draft_human_review_and_timed_pilot_pending", "identifier_policy": { "human_visible_namespaces": ["PAPER", "PLAT"], "paper_display_pattern": "PAPER-[0-9]{2}", "platform_display_pattern": "PLAT-[0-9]{2}", "internal_keys_are_metadata_only": true, "boundary": "Human-facing narrative uses stable display IDs; internal catalog rows or citation keys do not replace them." }, "required_artifacts": [ "README.md", "LECTURE_NOTES.md", "SLIDE_SCRIPT.md", "WORKED_CASE.md", "TRANSCRIPT.md", "manifest.json", "validate_package.py" ], "learning_outcomes": [ "Define an estimand with treatment, control, unit, outcome, population, state, contrast, weights, uptake, interference, and missingness.", "Distinguish potential outcomes, estimands, estimators, estimates, and transported claims.", "Draw and critique DAG paths for confounding, mediation, collider selection, drift, measurement change, and interference.", "Choose randomization, blocking, matching, staggered, switchback, difference-in-differences, or interrupted-series designs by assignment feasibility and failure mode.", "Plan power and uncertainty at the independent assignment unit while preserving clustered repeated measurements.", "Predeclare outcomes, multiplicity, missingness, retries, sequential monitoring, guardrails, and release rules.", "Maintain an append-only deviation log and downgrade causal language when identification conditions fail.", "Separate internal, construct, statistical-conclusion, and external validity.", "Reproduce L05, L06, and a bounded pair-blocked synthetic contrast without making a live-platform causal claim." ], "time_design": { "status": "authoring_budgets_not_observed_delivery_times", "synchronous_seminar_minutes": 90, "synchronous_core_studio_minutes": 75, "asynchronous_core_minutes_min": 200, "asynchronous_core_minutes_max": 220, "transcript_equivalent_minutes_min": 35, "transcript_equivalent_minutes_max": 45, "transcript_planned_minutes": 45, "pilot_state": "not_taught_not_rehearsed_not_timed", "media_state": "no_video_or_audio_recording_exists" }, "local_dependencies": [ { "id": "COURSE-W08", "path": "course-site/app/lib/course-data.ts#W08", "role": "Authoritative title, essential question, outcomes, prerequisites, paper routes, L05/L06 links, and evidence boundary.", "boundary": "A curriculum record does not establish lesson delivery, learning effectiveness, or causal validity." }, { "id": "NOTES-EXPERIMENTS", "path": "notes-latex/chapters/07-experiments.tex", "role": "Local alignment for treatment versions, potential outcomes, assignment units, temporal designs, missingness, interference, and validity.", "boundary": "W08 uses independently authored prose and visuals; Core Notes are not copied and do not identify a closed platform." }, { "id": "NOTES-CAUSAL-BRIDGE", "path": "notes-latex/appendices/f-prerequisite-bridges.tex#sec:bridge-potential-outcomes", "role": "Notation and prerequisite alignment for potential outcomes, estimands, repeated measures, and clustered designs.", "boundary": "A notation bridge states targets and assumptions; it does not make an observational comparison causal." }, { "id": "LAB-L05", "path": "course-labs/L05_controlled_content_intervention", "role": "Deterministic local treatment-version, factual-equivalence, accessibility-invariant, and rollback audit.", "boundary": "No system response or causal visibility effect is measured." }, { "id": "LAB-L06", "path": "course-labs/L06_repeated_measurement", "role": "Deterministic synthetic repeated-measurement panel and descriptive interval, drift, query-mix, and missingness outputs.", "boundary": "No intervention is assigned; repeated-query dependence remains; no named product or population effect is estimated." } ], "external_sources": [ { "display_id": "PAPER-10", "namespace": "PAPER", "catalog_row": 10, "internal_citation_key": "chen2025cc", "type": "arxiv_preprint", "title": "CC-GSEO-Bench: A Content-Centric Benchmark for Measuring Source Influence in Generative Search Engines", "authors": "Qiyuan Chen, Jiahe Chen, Hongsen Huang, Qian Shao, Jintai Chen, Renjie Hua, Hongxia Xu, Ruijia Wu, Ren Chuan, and Jian Wu", "doi": "10.48550/arXiv.2509.05607", "url": "https://arxiv.org/abs/2509.05607", "role": "Core bounded route for content-centric units, query clusters, aggregation, and distinct influence dimensions.", "boundary": "Benchmark construction, systems, query-article filtering, labels, and date constrain findings; it does not identify a universal source-influence effect." }, { "display_id": "PAPER-23", "namespace": "PAPER", "catalog_row": 23, "internal_citation_key": "kim2026sageo", "type": "arxiv_preprint", "title": "SAGEO Arena: A Realistic Environment for Evaluating Search-Augmented Generative Engine Optimization", "authors": "Sunghwan Kim, Wooseok Jeong, Serin Kim, Sangam Lee, and Dongha Lee", "doi": "10.48550/arXiv.2602.12187", "url": "https://arxiv.org/abs/2602.12187", "role": "Core bounded route for stage-aware evaluation in a reconstructed end-to-end pipeline.", "boundary": "Corpus, baseline target selection, pipeline implementation, systems, runs, and reporting define the ceiling; an open reconstruction does not disclose a closed product." }, { "display_id": "PAPER-32", "namespace": "PAPER", "catalog_row": 32, "internal_citation_key": "wu2026from", "type": "arxiv_preprint_audit_only", "title": "From Experience to Skill: Multi-Agent Generative Engine Optimization via Reusable Strategy Learning", "authors": "Beining Wu, Fuyou Mao, Jiong Lin, Cheng Yang, Jiaxuan Lu, Yifu Guo, Siyu Zhang, Yifan Wu, Ying Huang, and Fu Li", "doi": "10.48550/arXiv.2604.19516", "url": "https://arxiv.org/abs/2604.19516", "role": "Audit-only case for benchmark size, test-set selection, strategy adaptation, attribution construction, run uncertainty, and interference questions.", "boundary": "No reported headline effect enters W08 as a course finding without claim-level verification of assignment, sampling, uncertainty, and analysis." }, { "display_id": "PLAT-04", "namespace": "PLAT", "catalog_row": 4, "internal_catalog_id": "P04", "type": "official_platform_documentation", "title": "Introducing AI Performance in Bing Webmaster Tools Public Preview", "authors": "Microsoft Bing Webmaster Team", "url": "https://blogs.bing.com/webmaster/February-2026/Introducing-AI-Performance-in-Bing-Webmaster-Tools-Public-Preview", "role": "Bounded platform-interface case for separating named counts and trends from causal attribution.", "boundary": "Official preview documentation is authoritative only for its named Microsoft surfaces and checked date; it provides no randomized assignment, counterfactual effect, or cross-platform transport." } ], "l05_fixture_contract": { "lab_id": "L05", "synthetic": true, "offline": true, "deterministic": true, "locked_claim_count": 3, "source_identity_count": 3, "changed_factor_count": 1, "factor": "evidence_layout_proximity", "equivalence_check_count": 6, "expected_stdout": "L05 PASS: 3 locked claims, 1 changed factor, 0 error(s)", "input_sha256": { "control.html": "fff13e70c2218f8ea607f93128f5845f5d2582e6973badd469143c1b1fc18e78", "treatment.html": "af5877a1fad74e12118ac61ff62d96e110e643371bafb488e20d575aec55ac92", "control.css": "69f8bd330f589b84d309843be1e7ad95af19eac04c4e05da4e5409552182383d", "treatment.css": "8a576c5ad33fe52a794ba875a39479c878a55f0cbe8ebecb7e1f8f199cde951c", "locked_claims.json": "791f4ac66868f263994bef45a93cbf353c164cc1b3e452b571a341af83619f6a", "intervention_card.json": "26bf06ea1b21a1e4469e87a16d2d6e2a09fc19861e3b97ba39b1e3f4a0d28809", "factual_equivalence.csv": "f3397697225142cfcec267c037d123984326085d803868a4ad9b29a18bb80867" }, "claim_ceiling": "local structural equivalence and reversibility only; no system response or causal visibility effect" }, "l06_fixture_contract": { "lab_id": "L06", "synthetic": true, "offline": true, "deterministic": true, "event_count": 360, "complete_event_count": 354, "noncomplete_event_count": 6, "query_count": 20, "surface_count": 2, "time_block_count": 3, "repetition_count": 3, "expected_stdout": "L06 PASS: 360 events, 20 queries, 2 surfaces, 3 blocks", "input_sha256": { "panel.csv": "c013f9aee92085ed91ef32261ad2e1a98956f1cc3667cd0c658985fb4a7ab5d7" }, "dependence_warning": "Repeated events for one query are not independent; Wilson intervals are descriptive only.", "claim_ceiling": "no named product, live engine, causal intervention, business outcome, or population demand" }, "synthetic_identification_case": { "case_id": "W08-PAIR-DID-001", "status": "deterministic_arithmetic_teaching_fixture_not_external_effect", "assignment_unit": "query_intent_cluster", "assignment_design": "four frozen pairs with one treatment and one control cluster per pair", "opportunities_per_cluster_period": 20, "equal_cluster_weights": true, "records": [ {"pair": 1, "cluster": "Q-01", "assignment": "treatment", "baseline_successes": 8, "post_successes": 11}, {"pair": 1, "cluster": "Q-02", "assignment": "control", "baseline_successes": 8, "post_successes": 9}, {"pair": 2, "cluster": "Q-03", "assignment": "treatment", "baseline_successes": 10, "post_successes": 12}, {"pair": 2, "cluster": "Q-04", "assignment": "control", "baseline_successes": 10, "post_successes": 11}, {"pair": 3, "cluster": "Q-05", "assignment": "treatment", "baseline_successes": 6, "post_successes": 9}, {"pair": 3, "cluster": "Q-06", "assignment": "control", "baseline_successes": 6, "post_successes": 7}, {"pair": 4, "cluster": "Q-07", "assignment": "treatment", "baseline_successes": 12, "post_successes": 13}, {"pair": 4, "cluster": "Q-08", "assignment": "control", "baseline_successes": 12, "post_successes": 13} ], "expected_treated_mean_change": 0.1125, "expected_control_mean_change": 0.05, "expected_difference_in_differences": 0.0625, "required_identification_conditions": [ "assignment integrity", "verified treatment uptake", "parallel untreated counterfactual trend", "stable outcome measurement", "no cross-cluster interference", "declared missingness and retry policy", "stable synthetic system state" ], "claim_ceiling": "reproduced cluster arithmetic only; no live platform, named product, user population, or empirical GEO effect" }, "visual_policy": { "external_figure_reuse": false, "external_screenshot_reuse": false, "visuals": "original_instructional_diagram_specifications_only", "required_encodings": ["label", "shape_or_position", "line_style_or_pattern", "alt_text", "table_equivalent_for_quantitative_visuals"], "license_review_required_before_any_later_external_asset_reuse": true }, "claim_boundary": { "non_equivalences": [ "after-before change is not automatically a treatment effect", "an estimator is not an estimand", "repetition is not an independent assignment unit", "a DAG is not proof", "statistical significance is not identification", "internal validity is not external validity", "separately valid treatment and panel fixtures do not form a joined experiment" ], "not_evaluated": [ "live platform treatment uptake", "public crawling or indexing response", "production retrieval, ranking, generation, or citation effect", "named product comparison", "business outcome", "population demand", "cross-engine transport", "longitudinal platform stability", "individual user welfare", "legal or regulatory compliance" ] }, "quality_status": { "structural_validator": "passed_2026-08-24", "deterministic_fixture_alignment": "passed_2026-08-24", "scientific_human_review": "pending", "independent_reproduction": "pending", "accessibility_human_review": "pending", "timed_rehearsal": "pending", "classroom_pilot": "pending", "grading_calibration": "pending", "external_methodological_review": "pending" } }