{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp03_full320_20260801","cell_id":"invariant_mode_decomposition_design__ethnography_qualitative_methods","trajectory_id":"R","attempt_index":0,"archetype_slug":"invariant_mode_decomposition_design","domain_slug":"ethnography_qualitative_methods","decision":"CANDIDATE","problem_id":"single_code_analysis_misses_persistent_interaction_configurations","causal_lever_id":"coupled_transition_mode_guided_analysis","proposal":{"problem":"In longitudinal ethnographic corpora, analysts commonly examine codes or themes separately even though meaning-bearing configurations of roles, practices, and contexts may recur and change together across successive episodes. This can make frequent individual codes appear important while persistent, escalating, or reversing configurations remain hidden, distorting explanation and subsequent sampling. The problem exists independently of modal analysis whenever conclusions depend on coupled temporal patterns that single-code inspection fails to recover.","actors_substrate":["Ethnographers and qualitative analysts","Participants represented in fieldnotes, interviews, or observations","A longitudinal corpus segmented into ordered episodes","A documented coding scheme and analytic memos"],"observable_state":"Across adjacent episodes, combinations of coded practices, roles, contexts, and stated meanings recur or change together more consistently than individual-code summaries reveal; held-out sequences may exhibit repeatable coupled transitions.","consequence":"Analysts may mistake coordinate-level frequency for explanatory importance, overlook a low-frequency but persistent or escalating configuration, and direct theoretical sampling or interpretation toward the wrong phenomenon.","affected_objective":"Improve the traceability, temporal adequacy, and boundary-awareness of qualitative explanations while preserving discordant cases and contextual meaning.","structural_mapping":[{"archetype_element":"Transformation acting on coupled variables","domain_realization":"A locally estimated transition from one coded episode-state vector to the next within a defined site, period, and coding frame.","claim_kind":"HYPOTHESIS"},{"archetype_element":"Invariant directions","domain_realization":"Approximately recurring combinations of codes whose proportions persist, grow, decay, or reverse under the estimated transition.","claim_kind":"INFERENCE"},{"archetype_element":"Scalar modal response","domain_realization":"Estimated persistence or amplification of each configuration across successive episodes, treated as descriptive rather than causal.","claim_kind":"INFERENCE"},{"archetype_element":"Action-relevant mode","domain_realization":"A configuration whose persistence, consequence, or contribution to held-out reconstruction warrants closer case comparison or theoretical sampling.","claim_kind":"HYPOTHESIS"},{"archetype_element":"Residual visibility","domain_realization":"Unreconstructed episodes and structured coding residuals remain available as possible negative cases rather than being discarded as noise.","claim_kind":"CORPUS"},{"archetype_element":"Local validity","domain_realization":"Modes are trusted only for the site, period, segmentation, and coding scheme under which the transition was estimated.","claim_kind":"INFERENCE"}],"component_map":[{"component":"Transformation Scope","status":"adapted","domain_realization":"Adjacent-episode transitions within one bounded corpus and analytic period."},{"component":"State-Vector Definition","status":"adapted","domain_realization":"Documented code intensities or presences, with role and context indicators, for each episode."},{"component":"Invariant Mode Basis","status":"adapted","domain_realization":"Approximate coupled code configurations of the fitted transition operator."},{"component":"Modal Gain Spectrum","status":"adapted","domain_realization":"Estimated persistence, decay, reversal, or amplification coefficients with uncertainty."},{"component":"Dominant Mode Selection Rule","status":"adapted","domain_realization":"Retain modes meeting preregistered held-out fidelity or consequence thresholds, not merely high variance."},{"component":"Stable/Unstable Mode Partition","status":"adapted","domain_realization":"Classify configurations as decaying, persistent, oscillating, or apparently growing within the bounded window."},{"component":"Modal Intervention Map","status":"adapted","domain_realization":"Translate retained configurations into cases to revisit, contrasts to seek, or sampling questions; do not manipulate participants."},{"component":"Reconstruction Residual Check","status":"direct","domain_realization":"Reconstruct held-out episode states and inspect both residual magnitude and discordant-case structure."},{"component":"Mode Drift Monitor","status":"adapted","domain_realization":"Re-estimate across time blocks and flag rotations, reordered modes, or coding-frame changes."},{"component":"Interpretation Scope Contract","status":"direct","domain_realization":"State that modes are descriptive analytic constructs, not participant essences or causal laws."},{"component":"Mode-Coupling Register","status":"adapted","domain_realization":"Record near-degenerate, non-orthogonal, or interpretation-entangled configurations."},{"component":"Local Linearization Window","status":"adapted","domain_realization":"Specify the site, dates, episode grain, and coding regime within which approximation is evaluated."},{"component":"Spectral Gap Threshold","status":"adapted","domain_realization":"Require adequate separation and bootstrap stability before naming a dominant configuration."}],"mechanism_dispositions":[{"slug":"eigendecomposition_workflow","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Decompose the explicit square transition estimate; report conditioning and uncertainty.","counterfactual_removal":"No invariant configurations or persistence spectrum would be produced."},{"slug":"modal_sensitivity_sweep","disposition":"selected_supporting","contribution_type":"TEST_DESIGN","adaptation_or_rejection":"Use analytic perturbations of coding and mode weights, never participant manipulation.","counterfactual_removal":"Robustness of sampling implications and cross-mode effects would remain unknown."},{"slug":"modal_stability_analysis","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Classify only within the corpus window and avoid causal language.","counterfactual_removal":"Persistent or growing configurations could not be distinguished from momentary prevalence."},{"slug":"mode_shape_testing","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Physical excitation has no ethical or structural analogue here; observation is already encoded in the corpus.","counterfactual_removal":"No loss; residual and held-out tests provide empirical checks."},{"slug":"network_spectral_centrality_analysis","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Node importance is not the target problem and could improperly rank participants.","counterfactual_removal":"No loss to transition-configuration analysis."},{"slug":"power_iteration_probe","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Dominant-only recovery hides secondary and near-degenerate configurations.","counterfactual_removal":"No loss because the bounded operator permits fuller decomposition."},{"slug":"principal_component_analysis","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Variance directions do not represent temporal transition or persistence.","counterfactual_removal":"No loss; avoids substituting variance for dynamics."},{"slug":"reduced_order_model","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"A runnable predictive surrogate would overstate the aim and evidential maturity.","counterfactual_removal":"Interpretive targeting remains possible without simulation."},{"slug":"residual_reconstruction_test","disposition":"selected_load_bearing","contribution_type":"TEST_DESIGN","adaptation_or_rejection":"Evaluate held-out episodes and inspect structured residuals as negative cases.","counterfactual_removal":"Compression could erase consequential exceptions without detection."},{"slug":"singular_value_decomposition","disposition":"selected_supporting","contribution_type":"TEST_DESIGN","adaptation_or_rejection":"Use as a robustness comparison when the fitted operator is non-normal or ill-conditioned, not as invariant dynamics.","counterfactual_removal":"Fragile eigenmode claims would have one fewer diagnostic check."},{"slug":"spectral_decomposition_report","disposition":"selected_supporting","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"Pair every mode with provenance, uncertainty, excluded meanings, couplings, and negative cases.","counterfactual_removal":"Tentative numerical patterns could be laundered into authoritative cultural claims."},{"slug":"spectral_gap_monitor","disposition":"selected_supporting","contribution_type":"OPERATIONAL","adaptation_or_rejection":"Track mode separation and rotation across prespecified time blocks.","counterfactual_removal":"A once-valid simplification could silently persist after analytic regime drift."}],"causal_chain":["Define ordered episodes and meaning-bearing coded states within a fixed interpretive scope.","Estimate and validate a local transition operator rather than treating codes independently.","Decompose it to expose coupled configurations and their bounded persistence or amplification.","Select configurations using held-out fidelity, consequences, stability, and gap criteria.","Route selected configurations to case revisiting or theoretical sampling while routing residuals to negative-case analysis.","Compare resulting analytic discoveries and revisions with baseline and rival workflows."],"baseline":"Ordinary thematic coding followed by code-frequency, quotation, and informal co-occurrence review, with sampling decisions made from salient individual themes.","nearest_rival":"Sequence-aware manual case comparison combined with negative-case analysis and theoretical sampling, without fitting a modal transition model.","authority_safety":{"affected_parties":["Participants represented in the corpus","Researchers whose interpretations and sampling choices may change","Communities or organizations that could be characterized by resulting claims","Data stewards and ethics oversight bodies"],"decision_authority":"The research team controls analytic use only within participant consent, data-governance, and ethics-review constraints; the model has no authority to characterize people or initiate field action.","authorized_first_step":"Run a retrospective, read-only pilot on one already-authorized corpus: fit on earlier episodes, test on a held-out block, blind analysts to method labels when comparing baseline, rival, and modal outputs, and stop after assessing reconstruction, negative-case discovery, interpretive revisions, and mode stability.","excluded_actions":["Ranking or labeling individual participants","Changing field conditions to excite a mode","Collecting new data beyond existing consent","Publishing modal labels as cultural essences or causal findings","Automating theoretical conclusions or participant-facing decisions"],"halt_rollback":"Halt if held-out advantage is absent, modes are unstable across reasonable coding or segmentation choices, residuals concentrate in consequential groups, or participants could be harmed by interpretation. Roll back to the original corpus, coding record, and baseline analysis; retain the modal output only as an audited failed hypothesis."}},"negative_tests":{"strongest_counterevidence":"The apparent modes may be artifacts of coder conventions, episode segmentation, repeated interview prompts, or unequal observation density; independently coded and time-split analyses may fail to reproduce them.","analogy_break":"Ethnographic meaning is contextual, reflexive, and sometimes deliberately discontinuous; a numeric state vector may not preserve it, and an estimated linear operator need not represent a genuine transformation used by the social setting.","failure_condition":"No defensible episode state or locally stable transition can be specified, or results change qualitatively under reasonable coding, grain, or time-window choices.","problem_falsifier":"Held-out and independently coded data show no reproducible coupled temporal structure beyond what individual-code summaries and ordinary case comparison recover.","intervention_falsifier":"Even when reproducible coupled structure exists, modal guidance produces no improvement over the baseline or nearest rival in held-out reconstruction, discovery of consequential negative cases, or documented analytic revision.","risks":["Reifying mathematical modes as participant types or cultural laws","Encoding coder bias and observation schedules as structural dynamics","Suppressing rare but consequential accounts through compression","Retrospective overfitting and false dominance from a small spectral gap","Redirecting sampling away from participant priorities","Disclosure risk from sparse configurations that indirectly identify people"]},"null_rationale":null,"classification":{"candidate_kind":"DOMAIN_TRANSFER","prior_art_status":"UNSEARCHED","evidence_maturity":"HYPOTHESIS"},"revision_change_log":{"revision_kind":"ORIGINAL","prior_problem_id":null,"prior_causal_lever_id":null,"problem_changed":false,"causal_lever_changed":false,"conceptual_changes":[],"operational_changes":[],"repairs_addressed":[]},"confidence":0.74,"generator_notes":"The fit is conditional on ordered data, defensible state construction, and a locally stable transition estimate. The proposal treats modes as bounded descriptive aids to qualitative inquiry, not as causal or ontological facts."}