{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp03_full320_20260801","cell_id":"invariant_mode_decomposition_design__rhetoric","trajectory_id":"R","attempt_index":0,"archetype_slug":"invariant_mode_decomposition_design","domain_slug":"rhetoric","decision":"CANDIDATE","problem_id":"coupled_audience_backlash_hidden_by_aggregate_message_metrics","causal_lever_id":"damp_backlash_response_mode_through_combined_message_revision","proposal":{"problem":"Message teams often evaluate claims, frames, and identity cues separately or through aggregate approval. Yet combinations of these elements can move trust, comprehension, identity threat, and claim acceptance together. A message can therefore trigger repeatable subgroup backlash while aggregate metrics remain stable, causing coordinate-wise revisions to chase salient phrases instead of the coupled response pattern.","actors_substrate":["Audiences and audience subgroups exposed to deliberative messages","Writers, speakers, editors, and campaign communication teams","Rhetoric researchers or evaluators conducting consented response studies","Messages represented by controlled combinations of claims, frames, evidence, and identity cues"],"observable_state":"For each consented audience stratum, a pre-message and post-message vector of comprehension, claim acceptance, source trust, perceived identity threat, emotional response, and intended action, measured across controlled message variants and follow-up intervals.","consequence":"A weakly damped or amplifying audience-response pattern can produce backlash, distrust, or polarization that single-outcome and aggregate dashboards detect late or misattribute to one surface feature.","affected_objective":"Reduce avoidable backlash and loss of trust while preserving comprehension, truthful content, audience autonomy, and legitimate deliberation.","structural_mapping":[{"archetype_element":"Transformation acting on coupled variables","domain_realization":"A locally estimated map from pre-exposure audience state and controlled message features to post-exposure rhetorical response.","claim_kind":"HYPOTHESIS"},{"archetype_element":"Invariant or approximately invariant directions","domain_realization":"Repeatable combinations such as simultaneous identity threat, distrust, and rejection that recur across nearby message variants.","claim_kind":"HYPOTHESIS"},{"archetype_element":"Scalar modal response","domain_realization":"Estimated persistence, amplification, or damping of each combined response pattern across exposure or follow-up.","claim_kind":"INFERENCE"},{"archetype_element":"Action-relevant dominant mode","domain_realization":"A response pattern selected by its predicted contribution to backlash or comprehension rather than variance alone.","claim_kind":"INFERENCE"},{"archetype_element":"Intervention in modal coordinates","domain_realization":"Revise a combination of frame, evidence placement, and identity cue predicted to reduce the harmful pattern while retaining factual claims.","claim_kind":"HYPOTHESIS"},{"archetype_element":"Residual and regime governance","domain_realization":"Track unexplained subgroup responses, mode rotation, and failures outside the studied audience, issue, channel, and time window.","claim_kind":"INFERENCE"}],"component_map":[{"component":"Transformation Scope","status":"adapted","domain_realization":"One issue, channel, audience frame, and short pre/post exposure interval."},{"component":"State-Vector Definition","status":"direct","domain_realization":"Joint audience-response measures plus controlled message-feature coordinates."},{"component":"Invariant Mode Basis","status":"adapted","domain_realization":"Approximately recurring coupled response directions estimated within the pilot window."},{"component":"Modal Gain Spectrum","status":"adapted","domain_realization":"Estimated persistence or amplification of response directions, with uncertainty."},{"component":"Dominant Mode Selection Rule","status":"direct","domain_realization":"Retain modes exceeding preregistered outcome sensitivity and fidelity thresholds."},{"component":"Stable/Unstable Mode Partition","status":"adapted","domain_realization":"Classify response patterns as decaying, persistent, or growing only within observed follow-up."},{"component":"Modal Intervention Map","status":"adapted","domain_realization":"Map feasible combinations of message revisions to predicted modal changes."},{"component":"Reconstruction Residual Check","status":"direct","domain_realization":"Reconstruct held-out response vectors and inspect both error magnitude and structured subgroup residuals."},{"component":"Mode Drift Monitor","status":"direct","domain_realization":"Re-estimate modes across waves and flag rotation, reordering, or subgroup-specific drift."},{"component":"Interpretation Scope Contract","status":"direct","domain_realization":"State that modes are local descriptive or predictive constructs, not audience essences or universal causal traits."},{"component":"Mode-Coupling Register","status":"direct","domain_realization":"Record cross-effects and near-degenerate patterns that cannot be interpreted independently."},{"component":"Local Linearization Window","status":"adapted","domain_realization":"The tested issue, variants, audience sample, channel, dose, and follow-up duration."},{"component":"Spectral Gap Threshold","status":"direct","domain_realization":"A preregistered minimum separation and bootstrap stability requirement for treating retained modes as distinct."}],"mechanism_dispositions":[{"slug":"eigendecomposition_workflow","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Decompose the explicitly estimated local square response operator; report conditioning and avoid independent-mode claims if it is non-normal or defective.","counterfactual_removal":"Without a decomposition, the proposed coupled directions and gains are not identified."},{"slug":"modal_sensitivity_sweep","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Perturb feasible message-feature combinations in the fitted model and then in randomized variants to rank outcome leverage and cross-effects.","counterfactual_removal":"Mode size would be mistaken for intervention relevance, breaking the lever selection."},{"slug":"modal_stability_analysis","disposition":"selected_supporting","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"Use only for observed short follow-up dynamics; label extrapolation beyond that window invalid.","counterfactual_removal":"The study could still find modes, but could not responsibly distinguish transient from persistent backlash."},{"slug":"mode_shape_testing","disposition":"selected_load_bearing","contribution_type":"OPERATIONAL","adaptation_or_rejection":"Replace physical excitation with consented randomized message variants and repeated audience-response measurement to estimate or validate the local operator.","counterfactual_removal":"Absent an equivalent empirical identification step, the modes would rest on an assumed rather than observed transformation."},{"slug":"network_spectral_centrality_analysis","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Audience-response dimensions are not nodes whose dominant eigenvector entries warrant an importance ranking.","counterfactual_removal":"No material change; node centrality is not part of the causal chain."},{"slug":"power_iteration_probe","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"The bounded state is small enough for a full spectrum, and a dominant-only estimate would hide near-degenerate harmful modes.","counterfactual_removal":"No material change; full decomposition supplies the needed basis."},{"slug":"principal_component_analysis","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Variance directions may initialize diagnostics but do not represent the pre/post rhetorical transformation or intervention leverage.","counterfactual_removal":"No material change; covariance variance is not the target mechanism."},{"slug":"reduced_order_model","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"A runnable real-time surrogate is unnecessary for the bounded first test and could encourage unjustified deployment.","counterfactual_removal":"No material change; the pilot needs inference and validation, not real-time control."},{"slug":"residual_reconstruction_test","disposition":"selected_load_bearing","contribution_type":"TEST_DESIGN","adaptation_or_rejection":"Choose retained modes using held-out reconstruction and inspect structured residuals by subgroup and outcome consequence.","counterfactual_removal":"There would be no defensible stopping rule or check for omitted low-variance but harmful responses."},{"slug":"singular_value_decomposition","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"SVD is reserved as a diagnostic if the fitted map is non-normal; paired input-output directions would not establish invariant audience-response modes.","counterfactual_removal":"No material change to the proposed invariant-mode test."},{"slug":"spectral_decomposition_report","disposition":"selected_supporting","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"Publish uncertainty, permissible interpretations, subgroup residuals, couplings, and explicit prohibitions on essentializing audiences.","counterfactual_removal":"The analysis remains computable, but downstream readers lose the hard boundary against causal and identity overclaims."},{"slug":"spectral_gap_monitor","disposition":"selected_supporting","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"Track bootstrap gap stability and mode rotation across study waves; suspend use when either threshold fails.","counterfactual_removal":"A once-valid simplification could remain in use after its retained modes cease to be distinct or stable." }],"causal_chain":["Controlled message combinations generate joint changes in several audience-response dimensions.","A locally stable response operator may contain a coupled distrust–identity-threat–rejection direction that aggregate or coordinate-wise metrics obscure.","Empirical modal estimation and sensitivity testing identify which feasible combination of message features loads onto that direction.","A revision that reduces loading on the harmful direction should reduce backlash without deleting truthful claims.","Held-out residual, subgroup, gap, and drift checks determine whether that explanation remains usable."],"baseline":"Ordinary practice uses aggregate favorability or engagement, subgroup averages, and isolated phrase or frame A/B tests, followed by intuitive copy revision.","nearest_rival":"A preregistered multivariate regression or factorial message experiment that estimates feature-by-outcome effects directly without asserting invariant response modes.","authority_safety":{"affected_parties":["Consenting study participants and represented audience groups","People later exposed to any revised message","Message authors and sponsors","Researchers and ethics or institutional reviewers"],"decision_authority":"An ethics-approved research lead may run the pilot; message owners and affected-community reviewers jointly decide whether any revision advances beyond research.","authorized_first_step":"Run a preregistered, consented pilot on one non-emergency issue using a small set of truthful variants, estimate modes on a training split, and compare one modal-guided revision with the baseline and nearest rival on held-out participants.","excluded_actions":["No covert exposure, individualized persuasion, political microtargeting, or use of protected traits for targeting","No fabrication, suppression of material facts, deceptive identity cues, or optimization solely for compliance","No production deployment or audience labeling from the pilot","No repeated exposure beyond the consented protocol"],"halt_rollback":"Stop exposure and discard the intervention recommendation if distress exceeds the approved bound, truthful comprehension worsens, subgroup harm increases, held-out residual exceeds tolerance, the spectral gap is unstable, or modes drift beyond threshold; revert to unmodified truthful messaging and independent review."}},"negative_tests":{"strongest_counterevidence":"Rhetorical response may be too interpretive, reflexive, heterogeneous, and context-dependent for a stable local operator; apparent modes could be sampling artifacts or survey-method covariance.","analogy_break":"Unlike a fixed physical transformation, audiences learn, contest premises, communicate with one another, and alter the transformation when they recognize persuasive intent; invariant directions may dissolve across issues or occasions.","failure_condition":"No candidate mode is reproducible across resamples and held-out variants, or acceptable reconstruction requires nearly all original dimensions, leaving no stable decision-relevant decomposition.","problem_falsifier":"Coordinate-wise or direct factorial models predict subgroup backlash and comprehension as well as the modal model, while no repeatable coupled response pattern remains after measurement artifacts and subgroup composition are controlled.","intervention_falsifier":"Against equal-cost baseline and direct-regression revisions, the modal-guided revision fails to reduce preregistered backlash or preserve truthful comprehension on held-out consenting participants.","risks":["Reifying statistical modes as fixed audience identities","Optimizing persuasion at the expense of autonomy or truthful deliberation","Survey covariance or demand effects masquerading as structure","Aggregate improvements concealing subgroup harm","Local estimates being extrapolated to new issues, channels, or populations","Mode instability producing confident but arbitrary message revisions"]},"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":"Closed-book structural inference. Empirical existence, stability, and intervention value of rhetorical response modes remain hypotheses for the bounded pilot."}