{"schema_version":1,"research_id":"eoa_inverse_innovation_exp05_external_evaluation_20260803","source_assessment_id":"invariant_mode_decomposition_design__art_aesthetics:P4:v0","cell_id":"invariant_mode_decomposition_design__art_aesthetics","search_queries":["site:tate.org.uk collection acquisition policy trustees criteria acquisition works","site:si.edu collections management policy acquisition authority Smithsonian directive","museum acquisition committee policy criteria official art museum","site:aam-us.org collections stewardship acquisition policy ethics museums","social influence undermines wisdom of crowds experiment PNAS 2011 Lorenz primary","equality bias impairs collective intelligence groups experiment PNAS 2015 Mahmoodi","group deliberation opinion dynamics linear transformation eigenvalues social influence network theory primary research","Delphi method controlled feedback anonymity iteration RAND official","dynamic mode decomposition opinion dynamics group deliberation scores","linear influence model repeated opinions estimate influence matrix primary research social network DeGroot","Dynamic mode decomposition data snapshots exact DMD original paper Schmid 2010","committee decision making nominal group technique independent scoring discussion ranking official guidance","Network dynamics of social influence in the wisdom of crowds Becker Brackbill Centola 2017 PubMed","site:pubmed.ncbi.nlm.nih.gov \"Network dynamics of social influence\"","site:law.cornell.edu 45 CFR 46 human subject research definition identifiable private information","site:hhs.gov OHRP 45 CFR 46 research human subjects private information"],"sources":[{"source_id":"S1","title":"How social influence can undermine the wisdom of crowd effect","publisher":"Proceedings of the National Academy of Sciences / PubMed, U.S. National Library of Medicine","url":"https://pubmed.ncbi.nlm.nih.gov/21576485/","source_class":"PRIMARY_RESEARCH","publication_date":"2011-05-16","accessed_at":"2026-08-03","claims_supported":["A controlled experiment with 144 participants found that social information narrowed opinion diversity without reliably improving collective error.","Repeated exposure to others' estimates increased confidence after convergence despite no corresponding accuracy improvement.","Successive private estimates provide an established experimental design for measuring deliberative updating, although the tasks were factual estimates rather than art acquisitions."]},{"source_id":"S2","title":"Network dynamics of social influence in the wisdom of crowds","publisher":"Proceedings of the National Academy of Sciences / PubMed, U.S. National Library of Medicine","url":"https://pubmed.ncbi.nlm.nih.gov/28607070/","source_class":"PRIMARY_RESEARCH","publication_date":"2017-06-12","accessed_at":"2026-08-03","claims_supported":["Network structure and influence patterns can change the effect of social updating on group judgment.","Social influence does not have a uniformly harmful effect; decentralized interaction can improve group estimates under some conditions.","Observed convergence alone therefore cannot establish that deliberation is defective or that a checkpoint will help."]},{"source_id":"S3","title":"Dynamic mode decomposition with control","publisher":"arXiv","url":"https://arxiv.org/abs/1409.6358","source_class":"PRIMARY_RESEARCH","publication_date":"2014-09-22","accessed_at":"2026-08-03","claims_supported":["DMD with control already estimates low-order dynamics and intervention effects from paired state and control snapshots.","Separating intrinsic dynamics from procedural interventions requires explicit control-input data; ordinary decomposition can confound the two.","The proposal's update-matrix decomposition and procedural sensitivity map are technically close to established DMDc and system-identification practice."]},{"source_id":"S4","title":"Evaluation Briefs No. 7: Gaining Consensus Among Stakeholders Through the Nominal Group Technique","publisher":"U.S. Centers for Disease Control and Prevention","url":"https://www.cdc.gov/healthy-youth/php/program-evaluation/pdf/brief7.pdf","source_class":"OFFICIAL_GUIDANCE","publication_date":"n.d.","accessed_at":"2026-08-03","claims_supported":["Nominal Group Technique already combines independent contribution, structured discussion, private ranking, and aggregation.","The method is intended to balance individual influence, limit domination and conformity pressure, and support democratic prioritization.","NGT supplies a low-complexity procedural comparator but can restrict discussion and idea development."]},{"source_id":"S5","title":"Collections Management Policy","publisher":"American Alliance of Museums","url":"https://www.aam-us.org/programs/ethics-standards-and-professional-practices/collections-management-policy/","source_class":"STANDARD","publication_date":"n.d.","accessed_at":"2026-08-03","claims_supported":["Museum collections policies should specify acquisition criteria, decision-making authority, documentation, ethical issues, and responsibility for collection decisions.","A process checkpoint would require approval under the institution's governing collections policy rather than unilateral adoption by an analyst or facilitator.","The standard supports traceability and explicit authority but expresses no need for modal analytics."]},{"source_id":"S6","title":"Collections Stewardship Standards","publisher":"American Alliance of Museums","url":"https://www.aam-us.org/programs/ethics-standards-and-professional-practices/collections-stewardship-standards/","source_class":"STANDARD","publication_date":"n.d.","accessed_at":"2026-08-03","claims_supported":["Museums are expected to acquire and document collections legally, ethically, and responsibly under current approved policies.","Standards call for thoughtful decision-making, regular review of policies and procedures, adequate personnel, and allocation of resources.","These requirements create general demand for accountable acquisition processes, not demonstrated demand for monitoring score-update modes."]},{"source_id":"S7","title":"CM Policy and Procedures: Acquisitions","publisher":"University of Colorado Boulder Art Museum","url":"https://www.colorado.edu/cuartmuseum/about/strategic-guiding-documents/collection-management-policy-and-procedures/cm-policy-and-3","source_class":"OFFICIAL_ORGANIZATION_DATA","publication_date":"n.d.","accessed_at":"2026-08-03","claims_supported":["An identifiable art-museum adopter class exists: the director, curator, staff committee, and Collection Committee conduct acquisition review under declared criteria.","The museum requires a written acquisition proposal and justification, committee approval by a majority of a quorum, and minutes recording acceptance or rejection.","This policy identifies an authorizing workflow but contains no expressed concern about hidden coupled score dynamics and does not require criterion scoring or repeated rescoring."]},{"source_id":"S8","title":"45 CFR 46.102—Definitions for purposes of this policy","publisher":"Electronic Code of Federal Regulations, U.S. Office of the Federal Register","url":"https://www.ecfr.gov/current/title-45/subtitle-A/subchapter-A/part-46/subpart-A/section-46.102","source_class":"GOVERNMENT_OR_REGULATOR","publication_date":"current through 2026-07-30","accessed_at":"2026-08-03","claims_supported":["A systematic investigation designed to develop or contribute to generalizable knowledge can constitute research.","Research involving interaction with living individuals or identifiable private information can constitute human-subjects research.","Coded score and rationale data from a small committee may remain readily identifiable, so an authorized institutional determination is needed rather than assuming that coding resolves oversight obligations."]}],"problem_evidence":{"support":"MODERATE","rationale":"Controlled studies visibly establish that repeated social feedback can reduce diversity, increase confidence without accuracy, and produce network-dependent changes in collective judgment. Museum standards and an actual art-museum policy confirm that acquisition decisions are consequential, criteria-governed, documented committee acts. However, no source demonstrates that art-acquisition committees exhibit reproducible coupled criterion/dispersion/confidence modes, that ordinary summaries miss them, or that such modes cause defective acquisitions. Transfer from factual-estimation experiments to plural, evidence-responsive aesthetic judgment is a major unresolved gap.","source_ids":["S1","S2","S5","S6","S7"]},"stakeholder_evidence":{"support":"MODERATE","rationale":"The CU Art Museum's director, curator, staff committee, and Collection Committee are identifiable examples of actors able to authorize a nonbinding process study; AAM standards confirm that governing authority, criteria, records, ethics, and policy review matter across museums. This is credible institutional pull for accountable and documented deliberation, but no museum or funder was found expressing demand for modal decomposition, repeated private scoring, or an algorithmic checkpoint. Adopter willingness and perceived burden require direct inquiry.","source_ids":["S5","S6","S7"]},"prior_art":{"proximity":"ADJACENT_PRIOR_ART","closest_analogues":[{"name":"Dynamic Mode Decomposition with control","similarity":"Estimates a local state-transition model from paired snapshots, decomposes its dynamics, and estimates how explicit control inputs alter those dynamics—the proposal's core technical workflow.","remaining_difference":"No located source applies DMDc to art-acquisition deliberation or wraps it in a human-authorized checkpoint protecting dissent and interpretive limits.","source_ids":["S3"]},{"name":"Nominal Group Technique","similarity":"Uses independent input, structured discussion, private ranking, and facilitation to limit domination and conformity while producing a group priority.","remaining_difference":"NGT does not estimate invariant update modes, spectral gains, residuals, or intervention-specific modal effects; it is a procedural comparator rather than a diagnostic model.","source_ids":["S4"]},{"name":"Repeated-estimate social-influence experiments and DeGroot-style influence analysis","similarity":"Measure successive individual revisions after social information and analyze convergence, diversity, confidence, and network-dependent influence.","remaining_difference":"The experiments concern truth-evaluable factual estimates, not multi-criterion aesthetic acquisition, and do not deploy a governance trigger based on coupled criterion-level modes.","source_ids":["S1","S2"]},{"name":"Policy-governed art-museum acquisition committee","similarity":"Uses declared criteria, written justification, committee review, voting authority, and recorded decisions to govern acquisitions.","remaining_difference":"The located policy does not require private numerical scoring, repeated rounds, dynamic modeling, or modal process intervention.","source_ids":["S5","S6","S7"]}],"distinctive_claim_remaining":"For one consenting art-acquisition committee under a fixed rubric and nonlive practice cases, a regularized modal state-transition model will (a) predict held-out round-to-round criterion, dispersion, and confidence updates better than mean-reversion and independent criterion-wise models and (b) identify a reproducible process mode whose amplitude is selectively changed by a preapproved procedural safeguard without reducing evidence traceability or recorded disagreement. This is an empirical, falsifiable local claim, not a claim about artistic quality or superior acquisition choices.","confidence":"HIGH"},"implementation_evidence":{"support":"MODERATE","rationale":"The computation is implementable with standard linear algebra and DMDc-style system identification, and existing committee procedures can accommodate written justification, private input, facilitation, and recorded authority. The current one-session concept is nevertheless likely underpowered if the state includes separate means, dispersions, and confidence for many criteria: the transition matrix and control effects can become underdetermined, while ordinal scales, candidate-specific evidence, changing discussion content, nonlinearity, and near-degenerate modes threaten interpretation. Workflow risks include added burden, strategic scoring, chilled discussion, and false authority. Small-group coding may not prevent reidentification; institutional research/privacy review and explicit consent or another authorized basis are prerequisites. No legal conclusion is offered because institution, jurisdiction, employment rules, and intended dissemination are unspecified.","source_ids":["S1","S2","S3","S4","S5","S7","S8"]},"scores":{"meaningful_impact":{"score":3,"rationale":"Acquisitions create lasting stewardship obligations, and misleading convergence could affect public-trust decisions; neither prevalence nor decision harm in art committees is measured.","source_ids":["S5","S6","S7"]},"stakeholder_pull":{"score":2,"rationale":"Museums visibly require accountable, authorized, documented acquisition processes, but no expressed demand for this modal checkpoint was located.","source_ids":["S5","S6","S7"]},"incremental_advantage":{"score":2,"rationale":"The proposal could reveal coupled dynamics missed by criterion-wise summaries, but it must outperform simple change reports, mean reversion, structured facilitation, and NGT; no comparative evidence exists.","source_ids":["S3","S4"]},"distinctiveness_plausibility":{"score":3,"rationale":"The art-acquisition governance application and dissent-preserving modal trigger were not found, although the analytical and procedural components are established adjacent art.","source_ids":["S3","S4","S7"]},"technical_implementability":{"score":2,"rationale":"Software and mathematics are available, but adequate independent transitions, stationarity, scale validity, mode separation, and control identification are unresolved for a small committee.","source_ids":["S1","S2","S3"]},"adoption_authority_feasibility":{"score":3,"rationale":"A director and Collection Committee can plausibly approve a nonbinding calibration under policy, subject to governing-authority, privacy, and research-review requirements.","source_ids":["S5","S7","S8"]},"evidence_readiness":{"score":2,"rationale":"Generic social-influence evidence and technical prior art exist, but domain prevalence, stakeholder acceptance, parameter identifiability, predictive value, and intervention effects require proprietary field data.","source_ids":["S1","S2","S3","S7"]},"safety_net_benefit":{"score":4,"rationale":"The proposed safeguard is reversible, nonbinding, preserves human authority, prohibits artwork ranking, and can expose residual disagreement; privacy and mathematization risks prevent a maximum score.","source_ids":["S5","S7","S8"]},"scalability":{"score":2,"rationale":"Forms and analysis code could be reused, but each committee, rubric, membership change, and deliberation format requires fresh authorization and likely re-estimation, limiting low-cost diffusion.","source_ids":["S3","S5","S7"]}},"score_confidence":"MODERATE","costs":{"first_evidence":{"band_2026_usd":"10K_TO_50K","scope":"One partnered feasibility study: governance and research-status determination, consent/privacy design, preregistration, nonlive practice materials, secure score collection, facilitation, baseline modeling, and an initial held-out analysis.","confidence":"LOW","assumptions":["One committee of roughly 6–10 members.","Approximately 8–12 nonlive practice cases spread across several sessions.","About 150–400 resource-equivalent staff, participant, governance, data, and analysis hours.","No custom production software, artwork purchases, or participant travel.","Band is a planning estimate; no vendor quote or externally verified labor-rate source was collected."],"source_ids":["S3","S4","S7","S8"]},"initial_deployment_startup":{"band_2026_usd":"50K_TO_250K","scope":"Build a reusable secure workflow, obtain enough calibration transitions for a constrained model, validate scoring reliability, document the interpretation contract, train facilitators, and complete institutional approvals for one museum.","confidence":"LOW","assumptions":["Uses existing survey, secure-storage, and statistical-computing infrastructure.","Includes specialist statistical and privacy review.","Does not include integration with a museum collection-management platform.","Multiple calibration sessions are needed because a single session is unlikely to identify the model credibly."],"source_ids":["S3","S5","S7","S8"]},"operational_launch":{"band_2026_usd":"50K_TO_250K","scope":"Independent nonlive replication, comparator analysis, governance approval of any checkpoint threshold, operating procedures, audit logging, staff training, and rollback rehearsal before any live procedural use.","confidence":"LOW","assumptions":["Limited to one institution and one committee/rubric configuration.","No automated artwork recommendation or ranking is built.","External methodological review and committee time are included.","A failed replication ends launch rather than expanding scope."],"source_ids":["S3","S5","S6","S7","S8"]},"annual_recurring":{"band_2026_usd":"10K_TO_50K","scope":"Secure administration, periodic nonlive recalibration, drift and residual review, governance reporting, facilitator refresh, access review, and incident/rollback readiness for one committee.","confidence":"LOW","assumptions":["One or two review cycles per year.","No material rubric or membership change requiring a full redevelopment.","Data retention remains minimal and local.","Band is resource-equivalent and excludes acquisition budgets."],"source_ids":["S5","S6","S7","S8"]}},"verified_pipeline_gates":{"externally_supported_problem":{"status":"YES","reason":"Primary experiments support the general possibility that social updating can narrow diversity, inflate confidence, and depend on influence structure, while museum sources establish that acquisition deliberation matters. Domain-specific prevalence remains unverified but does not erase the externally supported general problem.","source_ids":["S1","S2","S5","S6","S7"]},"externally_credible_adopter_or_authorizer":{"status":"YES","reason":"The CU Art Museum policy identifies a director, curator, staff committee, Collection Committee, and provost-governed acquisition workflow capable of authorizing or rejecting a nonbinding calibration. No actual adoption interest has been established.","source_ids":["S5","S7"]},"distinct_testable_incremental_claim":{"status":"YES","reason":"Held-out transition prediction and selective procedural effects can be compared directly with mean-reversion, criterion-wise, and structured-facilitation baselines, with explicit stability and harm falsifiers.","source_ids":["S3","S4"]},"bounded_next_evidence_step":{"status":"YES","reason":"A single-institution, nonlive, preregistered feasibility study with fixed cases, case-level holdout, reversible controls, and stop rules is bounded and decision-relevant.","source_ids":["S3","S4","S7","S8"]},"no_unresolved_safety_or_authority_stop":{"status":"UNCERTAIN","reason":"Committee policy approval, voluntary participation, confidentiality, data-retention authority, and an institutional determination of whether the activity is human-subjects research are unresolved. The study must not begin until these are documented.","source_ids":["S5","S7","S8"]},"credible_cost_scope_and_range":{"status":"UNCERTAIN","reason":"The four bands have explicit scopes and resource assumptions, but no labor-rate source, institutional estimate, security assessment, or vendor quote was obtained; ranges are planning-level only.","source_ids":["S3","S5","S7","S8"]}},"next_evidence_step":"Partner with one university art museum and first obtain written director/Collection Committee authorization plus an institutional privacy and human-research determination. Preregister a nonlive feasibility study with 6–10 consenting members, a fixed rubric, 8–12 authorized practice cases, and repeated private score–discussion–rescore transitions across several sessions. Before data collection, cap or regularize the state dimension according to a parameter-identifiability audit. Randomize a reversible evidence-first/private-rescore safeguard at the practice-case level. Split whole cases—not individual rows—into development and holdout sets. Compare a regularized DMDc/modal model against (1) no-change, (2) mean-reversion, (3) independent criterion-wise autoregression, and (4) an NGT-like structured process without a modal trigger. Advance only if the modal model improves held-out error by a preregistered practically meaningful margin, retained directions remain aligned across case-level bootstrap samples, the retained/dropped subspaces remain separated, and the safeguard changes the targeted mode without reducing evidence-linked explanations, recorded disagreement, participation, or perceived psychological safety. Falsify or retire the approach if simple baselines match it, modes rotate or merge, residuals remain structured, results depend on one case/member, participants report coercion or chilling, reidentification controls fail, or the procedural variation has diffuse, adverse, or irreproducible effects. The result may authorize only another nonlive study, not a live acquisition trigger.","blocking_evidence":["No direct evidence establishes the prevalence or consequences of hidden coupled update modes in art-acquisition committees.","No museum, committee, insurer, accreditor, or funder has expressed demand for modal deliberation monitoring.","The minimum stable state dimension and required number of independent case-round transitions are unknown.","Practice-case dynamics may not transfer to live, high-stakes acquisition decisions.","The validity of treating rubric scores and confidence scales as a locally linear state is untested.","No comparative data show advantage over criterion-wise change reports, mean reversion, NGT, or skilled facilitation.","Committee approval, participant voluntariness, privacy controls, retention rules, and human-research status remain undetermined.","Cost bands lack institution-specific labor rates, security estimates, and implementation quotes."],"research_disposition":"PARTNERED_RESEARCH_PROGRAM","world_novelty_boundary":"World novelty, patentability, freedom to operate, market size, and realized impact were not measured. This eight-source search found established social-influence experiments, DMDc/system-identification methods, structured consensus practice, museum standards, and a real acquisition-authority workflow, but it was not an exhaustive scholarly, product, procurement, or patent search. The only defensible remaining distinction is the unvalidated synthesis of those elements into a dissent-preserving modal checkpoint for art-acquisition deliberation.","arm":"COMPLETE_PROPOSAL_PORTFOLIO","candidate_version":0,"controller_recommendation":{"action":"STOP_EMPIRICAL_RESEARCH_NEEDED","repairable":false,"material_progress_observed":true,"progress_targets":["Secure a named museum partner and written authorization from the director and relevant collection/governing committee.","Obtain an institutional privacy and human-research determination before collecting coded private scores or rationales.","Complete a parameter-identifiability audit that fixes the state dimension, regularization, minimum transition count, and holdout plan before fitting modes.","Collect nonlive, case-held-out transition data and compare against no-change, mean-reversion, criterion-wise, and NGT-like procedural baselines.","Demonstrate bootstrap-stable modal directions, adequate subspace separation, low unstructured holdout residuals, and no single-case or single-member dependence.","Show that a randomized reversible safeguard selectively changes the preregistered mode without suppressing disagreement, traceability, participation, or psychological safety.","Obtain institution-specific resource estimates and data-security requirements before considering operational launch."],"reason":"Web evidence establishes a plausible generic problem, credible authorizers, close analytical and procedural prior art, and a bounded study design. It cannot establish that the proposed modes exist in art-acquisition deliberation, are identifiable from feasible samples, outperform simpler procedures, or respond safely to a checkpoint. Those questions require consented proprietary committee data and live nonbinding testing, so the next decision is empirical rather than additional bounded web research."},"proposal_index":4}