{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp03_full320_20260801","cell_id":"invariant_mode_decomposition_design__behavioral_economics","trajectory_id":"R","attempt_index":0,"archetype_slug":"invariant_mode_decomposition_design","domain_slug":"behavioral_economics","decision":"CANDIDATE","problem_id":"popularity_signal_cascade_hidden_mode","causal_lever_id":"mode_targeted_social_signal_damping","proposal":{"problem":"A marketplace that displays recent popularity can have stable aggregate conversion while coupled, sequential imitation across products or buyer groups amplifies arbitrary early choices into persistent demand cascades. Item-by-item metrics may miss the growing combination of choice shares that suppresses private quality information and misallocates attention.","actors_substrate":["buyers making sequential choices under uncertainty","sellers whose visibility and demand are affected","marketplace ranking and experimentation teams","consumer-protection or research reviewers"],"observable_state":"At each interval, a vector containing product-by-group choice shares, displayed popularity ranks, exposure shares, and available private-quality proxies; its lagged transition and interventions are observable within a prespecified market window.","consequence":"A self-reinforcing imitation pattern can concentrate demand on initially popular products without corresponding quality evidence, disadvantaging buyers and sellers even when aggregate conversion remains stable. This consequence is a HYPOTHESIS to be tested.","affected_objective":"Preserve choice quality and seller opportunity while limiting popularity-driven cascades without removing useful social information indiscriminately.","structural_mapping":[{"archetype_element":"coupled transformation","domain_realization":"The estimated lagged map from current exposure, displayed popularity, and group-level choices to the next interval's choice-share vector.","claim_kind":"HYPOTHESIS"},{"archetype_element":"invariant direction","domain_realization":"A recurring weighted combination of products and buyer groups whose pattern approximately persists while its magnitude changes.","claim_kind":"INFERENCE"},{"archetype_element":"modal gain","domain_realization":"The locally estimated persistence or amplification of that behavioral combination across decision intervals.","claim_kind":"INFERENCE"},{"archetype_element":"unstable hidden mode","domain_realization":"A cross-product imitation pattern that grows while aggregate conversion or individual product metrics appear acceptable.","claim_kind":"HYPOTHESIS"},{"archetype_element":"modal control","domain_realization":"Delay, cap, or reduce popularity-display strength only for exposures loading strongly on a validated cascade mode.","claim_kind":"HYPOTHESIS"}],"component_map":[{"component":"Transformation Scope","status":"adapted","domain_realization":"A locally stationary lagged transition for defined products, buyer groups, and exposure conditions."},{"component":"State-Vector Definition","status":"adapted","domain_realization":"Choice shares, ranks, exposures, and quality proxies indexed by product and buyer group."},{"component":"Invariant Mode Basis","status":"adapted","domain_realization":"Approximate eigenvectors of the estimated behavioral transition, traced back to their product-group weights."},{"component":"Modal Gain Spectrum","status":"adapted","domain_realization":"Estimated eigenvalues with uncertainty, supplemented by transient-gain diagnostics where non-normality matters."},{"component":"Dominant Mode Selection Rule","status":"adapted","domain_realization":"Retain modes exceeding preregistered persistence, outcome-sensitivity, and reconstruction criteria."},{"component":"Stable/Unstable Mode Partition","status":"adapted","domain_realization":"Classify locally decaying, marginal, oscillatory, and amplifying choice-share modes."},{"component":"Modal Intervention Map","status":"adapted","domain_realization":"Map feasible popularity-display changes to predicted changes in selected modal coordinates and buyer outcomes."},{"component":"Reconstruction Residual Check","status":"direct","domain_realization":"Out-of-sample error and structured residuals between observed states and retained-mode reconstructions."},{"component":"Mode Drift Monitor","status":"direct","domain_realization":"Scheduled checks for rotating modes, reordered gains, residual growth, and changed product topology."},{"component":"Interpretation Scope Contract","status":"direct","domain_realization":"Modes are local descriptive dynamics, not psychological traits or causal effects absent randomization."},{"component":"Mode-Coupling Register","status":"adapted","domain_realization":"Record near-degeneracy, non-normal cross-talk, and intervention spillovers among behavioral modes."},{"component":"Local Linearization Window","status":"adapted","domain_realization":"Fixed ranges of assortment, ranking policy, season, and exposure within which the transition approximation is evaluated."},{"component":"Spectral Gap Threshold","status":"direct","domain_realization":"A preregistered minimum separation, including uncertainty, between retained and discarded modes."}],"mechanism_dispositions":[{"slug":"eigendecomposition_workflow","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Decompose the explicitly estimated local lag-transition operator; require conditioning diagnostics.","counterfactual_removal":"Without it, persistent behavioral combinations and their gains are not identified."},{"slug":"modal_sensitivity_sweep","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Perturb simulated modal coordinates and feasible display controls, then confirm causal effects experimentally.","counterfactual_removal":"The analysis could rank persistent modes but could not identify actionable leverage or spillovers."},{"slug":"modal_stability_analysis","disposition":"selected_supporting","contribution_type":"TEST_DESIGN","adaptation_or_rejection":"Classify growth only inside the prespecified behavioral regime and attach uncertainty.","counterfactual_removal":"There would be no principled distinction between fading variation and a candidate cascade."},{"slug":"mode_shape_testing","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Physical excitation and sensor-based mode recovery do not transfer cleanly to autonomous human choice; randomized display tests serve validation instead.","counterfactual_removal":"No material change; behavioral validation remains available."},{"slug":"network_spectral_centrality_analysis","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Node centrality ranks products or users, whereas the problem concerns dynamic combinations and their gains.","counterfactual_removal":"No material change to cascade-mode identification."},{"slug":"power_iteration_probe","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"A dominant-only estimate could conceal a close competing mode and is unnecessary for the bounded state dimension.","counterfactual_removal":"No material change because the full estimated spectrum is used."},{"slug":"principal_component_analysis","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Variance directions need not represent lagged behavioral propagation or intervention leverage.","counterfactual_removal":"No material change; covariance compression is not the causal model."},{"slug":"reduced_order_model","disposition":"selected_supporting","contribution_type":"OPERATIONAL","adaptation_or_rejection":"Use a small local surrogate only for preregistered intervention simulations, never as evidence of causality.","counterfactual_removal":"Candidate controls would require costlier full-state simulation, but identification would remain."},{"slug":"residual_reconstruction_test","disposition":"selected_load_bearing","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"Test held-out reconstruction and residual structure, including quality-related low-variance behavior.","counterfactual_removal":"Omitted behavior could be mistaken for a trustworthy modal simplification."},{"slug":"singular_value_decomposition","disposition":"selected_supporting","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"Use SVD to diagnose conditioning and transient amplification, not to relabel singular vectors as invariant behavioral modes.","counterfactual_removal":"A non-normal fitted operator could yield a fragile eigenvector interpretation without warning."},{"slug":"spectral_decomposition_report","disposition":"selected_supporting","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"Document variable loadings, uncertainty, coupling, causal limits, and authorized uses.","counterfactual_removal":"Decision-makers could misread descriptive modes as psychological traits or established causes."},{"slug":"spectral_gap_monitor","disposition":"selected_load_bearing","contribution_type":"OPERATIONAL","adaptation_or_rejection":"Track retained-versus-discarded separation, subspace rotation, and residual drift during the pilot.","counterfactual_removal":"The targeting rule could continue after its modal basis loses identifiability."}],"causal_chain":["Visible popularity enters later buyers' information sets.","If imitation responses are coupled, repeated choice transitions produce approximately persistent product-group combinations.","A combination with validated gain near or above the stability boundary can amplify early choice noise while aggregate metrics conceal it.","Randomly reducing popularity-signal strength for exposures loading on that mode should reduce its subsequent amplitude if the display contributes causally.","Lower modal amplitude should preserve more responsiveness to private-quality proxies and reduce arbitrary demand concentration; both links are HYPOTHESES."],"baseline":"Monitor aggregate conversion and product-level trends, then adjust popularity badges or rankings one product at a time.","nearest_rival":"A conventional randomized trial that uniformly removes popularity information across all eligible exposures, without estimating coupled behavioral modes.","authority_safety":{"affected_parties":["buyers receiving altered social information","sellers gaining or losing exposure","marketplace operators","research and consumer-protection reviewers"],"decision_authority":"The marketplace experimentation owner, subject to independent ethics or consumer-protection review and seller-impact constraints.","authorized_first_step":"Run retrospective estimation followed by a shadow prospective test over one fixed assortment and time window; if preregistered fit and gap gates pass, randomize only a reversible display-strength reduction on a small eligible exposure fraction against uniform reduction and status quo.","excluded_actions":["targeting protected or psychologically inferred traits","fabricating popularity counts","changing prices or inventory","seller removal or punitive ranking","deploying outside the validated window","using descriptive modes as individual psychological labels"],"halt_rollback":"Stop and restore the original display when buyer-harm, seller-disparity, residual-error, mode-drift, or spectral-gap thresholds are crossed; invalidate the targeting rule pending re-estimation and review."}},"negative_tests":{"strongest_counterevidence":"Held-out transitions may be explained by independent product quality, inventory, marketing, or seasonality, with no stable coupled mode after those factors and uncertainty are included.","analogy_break":"Human choices are discrete, strategic, and policy-reactive; the transition is endogenous and nonlinear, so eigenmodes may rotate after a display change and need not be invariant beyond a narrow window.","failure_condition":"The proposal fails if no reproducible mode has adequate reconstruction, separation, temporal stability, and traceable product-group meaning, or if the fitted operator is too ill-conditioned for actionable interpretation.","problem_falsifier":"After controlling through the packet-authorized state representation for exposures and quality proxies, popularity information does not predict suppression of private-quality responsiveness or cross-product propagation beyond ordinary coordinate-level effects.","intervention_falsifier":"A properly powered randomized reduction of popularity-signal strength on high-loading exposures does not reduce the preregistered cascade-mode amplitude or improve quality responsiveness relative to status quo and uniform reduction.","risks":["A mode may encode omitted marketing, inventory, or demographic structure rather than imitation.","Targeted damping may redistribute seller exposure unfairly.","Small gaps or non-normality may make mode identity unstable.","Aggregate welfare gains may conceal subgroup harm.","Monitoring itself may normalize an unjustified causal interpretation."]},"null_rationale":null,"classification":{"candidate_kind":"MECHANISM_ADAPTATION","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":0.73,"generator_notes":"The fit depends on estimating a locally valid lagged choice-transition operator and validating the popularity display causally. Modal findings alone remain descriptive; prior art was not searched."}