{"schema_version":1,"research_id":"eoa_inverse_innovation_exp03_external48_20260801","source_assessment_id":"eoa_inverse_innovation_exp03_opportunity320_20260801","cell_id":"computability_boundary_mapping__behavioral_economics","selection_stratum":"REJECTION_LOW_BAND_AUDIT","search_queries":["behavioral economics agent based model formal verification reachability undecidable adaptive choice model","policy simulation behavioral models exact prediction limitations agent based official guidance","undecidability verification agent based models reachability Turing complete","formal verification cognitive models model checking behavioral economics","computational undecidability agent-based social systems prediction paper","behavioral policy agent based simulation guidance limitations prediction official OECD","formal methods cognitive architectures verification decidability paper","NIST AI RMF limitations uncertainty documentation intended use model outputs official","BLS occupational employment wage software developers economists 2025","ACT-R model checking formal verification reachability cognitive architecture"],"sources":[{"source_id":"S1","title":"Using Agent-Based Models for Prediction in Complex and Wicked Systems","publisher":"Journal of Artificial Societies and Social Simulation","url":"https://www.jasss.org/24/3/2.html","source_class":"PRIMARY_RESEARCH","publication_date":"2021-06-30","accessed_at":"2026-08-02","claims_supported":["Policy-relevant agent-based modelers run empirically calibrated models under alternative scenarios and must communicate the epistemic status of results.","The paper's practical definition of prediction permits uncertainty, imprecision, and occasional error rather than requiring an exact class-wide verdict.","For the finite empirical model-search problems considered, the authors characterize prediction as intractable or radically uncertain rather than formally undecidable."]},{"source_id":"S2","title":"Formal Verification of Cognitive Models","publisher":"AAAI Press / FLAIRS Conference","url":"https://cdn.aaai.org/FLAIRS/2006/Flairs06-082.pdf","source_class":"PRIMARY_RESEARCH","publication_date":"2006","accessed_at":"2026-08-02","claims_supported":["Automated formal verification of computational cognitive models was proposed using a specification language and algorithms for checking model correctness.","The approach distinguishes the model's formal traces from claims about simulated human behavior and acknowledges that cognitive tasks can be ill-structured.","The prototype automated construction and checking of a production-flow representation."]},{"source_id":"S3","title":"Formal verification of neural agents in non-deterministic environments","publisher":"Autonomous Agents and Multi-Agent Systems / Springer Nature","url":"https://doi.org/10.1007/s10458-021-09529-3","source_class":"PRIMARY_RESEARCH","publication_date":"2021-11-09","accessed_at":"2026-08-02","claims_supported":["Reachability verification for the paper's unbounded neural-agent/environment model is undecidable by an explicit reduction from the halting problem.","The authors recover a decidable bounded temporal-logic fragment, implement sequential and parallel MILP-based procedures, and evaluate them on examples.","The paper demonstrates that an undecidability boundary plus a bounded implemented fallback already exists in adjacent agent-verification research."]},{"source_id":"S4","title":"On Formal Verification of ACT-R Architectures and Models","publisher":"Cognitive Science Society / eScholarship","url":"https://escholarship.org/uc/item/1190b539","source_class":"PRIMARY_RESEARCH","publication_date":"2019","accessed_at":"2026-08-02","claims_supported":["ACT-R cognitive models have been given a formal operational model and translated into timed automata for automated defect analysis.","The work applied the translation to artificial models and a published cognitive model and examined scalability.","Formal verification of a behavior-predicting cognitive architecture therefore predates this candidate."]},{"source_id":"S5","title":"LOGIC: Good Practice Principles for Mainstreaming Behavioural Public Policy","publisher":"OECD","url":"https://www.oecd.org/en/publications/logic-good-practice-principles-for-mainstreaming-behavioural-public-policy_6cb52de2-en/full-report/component-5.html","source_class":"OFFICIAL_GUIDANCE","publication_date":"2024","accessed_at":"2026-08-02","claims_supported":["Behavioral-policy practice is framed as evidence-informed work involving scientific investigation, evaluation, contextual judgment, and sometimes new testing.","OECD reports that government behavioral-science teams exist and are maturing, while many still struggle to achieve implementation and impact.","The guidance does not specify demand for universal exact terminating behavioral-outcome prediction."]},{"source_id":"S6","title":"Good Practice Principles for Ethical Behavioural Science in Public Policy","publisher":"OECD","url":"https://doi.org/10.1787/e19a9be9-en","source_class":"OFFICIAL_GUIDANCE","publication_date":"2022-10-05","accessed_at":"2026-08-02","claims_supported":["Ethical considerations apply throughout behavioral-policy work from scoping through scaling.","Practitioners and policy makers are recognizable governance stakeholders for responsible behavioral-science use.","The guidance supports accountable review but does not identify an adopter for the proposed computability router."]},{"source_id":"S7","title":"AI Risk Management Framework Core","publisher":"National Institute of Standards and Technology","url":"https://airc.nist.gov/airmf-resources/airmf/5-sec-core/","source_class":"OFFICIAL_GUIDANCE","publication_date":"2023","accessed_at":"2026-08-02","claims_supported":["NIST calls for documenting intended scope, knowledge limits, human oversight, uncertainty, and limitations on generalizability.","NIST recommends independent assessment, pre-deployment testing, comparison with benchmarks, and safe failure beyond knowledge limits.","These governance practices are close analogues to explicit out-of-scope and UNKNOWN handling, although they are not a computability-specific behavioral-model implementation."]},{"source_id":"S8","title":"Test, Learn, Adapt: Developing Public Policy with Randomised Controlled Trials","publisher":"UK Cabinet Office and Behavioural Insights Team","url":"https://www.gov.uk/government/publications/test-learn-adapt-developing-public-policy-with-randomised-controlled-trials","source_class":"OFFICIAL_GUIDANCE","publication_date":"2012-06-14","accessed_at":"2026-08-02","claims_supported":["Official behavioral-policy guidance emphasizes predefined outcomes and comparison of interventions against a control condition.","This provides evidence that practical policy requirements are commonly empirical and comparative rather than universal exact model-level predictions."]}] ,"problem_evidence":{"support":"NOT_FOUND","rationale":"The search found a real adjacent concern—policy-facing simulations require careful communication of epistemic status—but no specification, procurement, product claim, or practitioner request for a correct, terminating verdict over every finitely described unrestricted adaptive behavioral model. The closest domain evidence instead defines useful prediction as uncertain and occasionally wrong and official practice emphasizes experiments, evaluation, and contextual evidence. This does not prove that no such demand exists, but the candidate's asserted problem incidence remains unverified.","source_ids":["S1","S5","S8"]},"stakeholder_evidence":{"support":"WEAK","rationale":"OECD guidance establishes behavioral-policy practitioners and policy makers as real actors with ethical and evidence-governance responsibilities, while NIST supplies a credible general governance role for independent assessment and scope documentation. No named organization or review owner expressed interest in adopting a computability-boundary checker or accepting the proposed joint behavioral/formal authority arrangement.","source_ids":["S5","S6","S7"]},"prior_art":{"proximity":"SUBSTANTIAL_COLLISION","closest_analogues":[{"name":"Bounded verification of neural agents after an undecidability result","similarity":"This work proves unbounded reachability verification undecidable by halting reduction, defines a bounded decidable fragment, implements analyzers, and evaluates them—the central technical pattern of the candidate.","remaining_difference":"Its subject is neural agents in nondeterministic environments, not an explicitly behavioral-policy choice-architecture language; it does not implement the proposed exact/bounded/UNKNOWN policy-facing router or test propagation of model-level labels into policy reports.","source_ids":["S3"]},{"name":"Formal verification of ACT-R cognitive models","similarity":"It formalizes a cognitive architecture, translates behavioral models into timed automata, and performs automated exhaustive defect analysis, directly occupying the cognitive-model verification portion of the proposed transfer.","remaining_difference":"It focuses on correctness and defects in ACT-R models rather than proving a universal reachability requirement undecidable, enforcing a language-class boundary, or producing an audited UNKNOWN outcome.","source_ids":["S4"]},{"name":"Formal Verification of Cognitive Models","similarity":"It proposes an automated specification-and-checking methodology for cognitive models and separates formal model behavior from human-like behavioral requirements.","remaining_difference":"It does not identify a halting-preserving behavioral-language encoding, a decidability router, or a policy-governance workflow.","source_ids":["S2"]},{"name":"NIST AI RMF scope and knowledge-limit governance","similarity":"It already recommends explicit scope, documented knowledge limits, uncertainty reporting, independent assessment, human oversight, and safe failure outside knowledge limits.","remaining_difference":"It is general risk guidance, not an executable formal-language membership checker or proof-backed behavioral-model analyzer.","source_ids":["S7"]}],"distinctive_claim_remaining":"The remaining testable claim is narrow: for one specifically declared behavioral-policy modeling language, a checked reduction establishes an unrestricted reachability boundary, while a machine-enforced router assigns every submitted model to a proved exact fragment, an explicitly horizon-bounded analysis, or UNKNOWN; the labels remain intact in downstream policy reports and the supported fragments cover enough representative models to be useful. No reviewed source demonstrated this complete behavioral-policy-specific combination.","confidence":"HIGH"},"implementation_evidence":{"support":"MODERATE","rationale":"Adjacent research demonstrates every major technical ingredient separately: formal semantics and automated checking for cognitive models, translation of ACT-R to timed automata, and a halting reduction followed by implemented bounded verification for neural-agent systems. NIST supplies mature governance patterns for scope, independent review, and safe failure. Implementability remains only moderate because the candidate has not declared its behavioral language, exhibited its reduction, defined enforceable fragment membership, or measured useful coverage.","source_ids":["S2","S3","S4","S7"]},"scores":{"meaningful_impact":{"score":2,"rationale":"Preventing definitive policy claims from unsupported model outputs could matter, but no evidence establishes that universal exact predictors are requested or causing harm. Existing policy-model practice already emphasizes uncertainty and empirical validation.","source_ids":["S1","S5","S8"]},"stakeholder_pull":{"score":1,"rationale":"Real behavioral-policy and governance stakeholders exist, but the bounded search found no adopter request, specification, commitment, or procurement for this computability-boundary workflow.","source_ids":["S5","S6","S7"]},"incremental_advantage":{"score":2,"rationale":"Explicit routing and preserved UNKNOWN could improve an undocumented binary-timeout workflow, but the technical core—undecidability analysis followed by bounded verification—and the governance core—scope and knowledge-limit documentation—already have close precedents. No baseline comparison demonstrates added coverage or fewer misinterpretations.","source_ids":["S3","S7"]},"distinctiveness_plausibility":{"score":1,"rationale":"The nearest agent-verification work substantially collides with the central technical design, and cognitive-model formal verification is established prior art. Only the integrated behavioral-policy application and downstream label audit remain distinguishable.","source_ids":["S2","S3","S4","S7"]},"technical_implementability":{"score":3,"rationale":"The proposed sandbox work is feasible in principle because comparable reductions, bounded analyzers, cognitive-model translations, and automated checks have been implemented. Feasibility for the unspecified target language and useful fragment coverage remains unproved.","source_ids":["S2","S3","S4"]},"adoption_authority_feasibility":{"score":2,"rationale":"Official guidance supports interdisciplinary governance, independent assessment, human oversight, and ethical review, so an authority structure is conceivable. No actual behavioral-policy organization has accepted the proposed joint owner or agreed that this formal analysis belongs in its workflow.","source_ids":["S6","S7"]},"evidence_readiness":{"score":2,"rationale":"Strong adjacent prior art makes a prototype research design possible, but the two threshold facts—an actual universal requirement and a declared language supporting the reduction—are absent.","source_ids":["S1","S3","S4","S5"]},"safety_net_benefit":{"score":4,"rationale":"Scope checks, explicit UNKNOWN, independent review, and safe failure beyond knowledge limits closely align with official risk-management guidance and would reduce overclaiming if the target workflow exists. The score does not imply that the formal result predicts human outcomes.","source_ids":["S7"]},"scalability":{"score":2,"rationale":"The governance pattern is reusable, but formal semantics, reductions, membership checkers, and tractable fragments are likely language-specific; adjacent work also treats scalability as a central concern.","source_ids":["S3","S4"]}},"score_confidence":"MODERATE","costs":{"first_evidence":{"band_2026_usd":"50K_TO_250K","scope":"A six- to sixteen-week, non-deployment study covering a demand audit, formal semantics for one language, one reduction attempt with independent proof review, a minimal membership checker, synthetic corpus creation, software/equipment, data-rights and research-governance review, coordination, reproducibility, and evaluation against finite-horizon and probabilistic baselines.","confidence":"MODERATE","assumptions":["Two to four fractional specialists cover behavioral modeling, formal methods, software implementation, and independent review.","Only synthetic or public model artifacts are used; no sensitive personal data or live intervention is included.","Existing solvers and ordinary cloud or workstation resources are sufficient.","The 2026 BLS wage data for software, mathematical, and operations-research occupations are labor-cost anchors; benefits, overhead, contracting premiums, and coordination raise resource-equivalent cost above direct wages."],"source_ids":["S3","S4","S7","S8"]},"initial_deployment_startup":{"band_2026_usd":"250K_TO_1M","scope":"If first evidence passes, engineer one governed analysis service with a versioned parser and semantics, proof and fragment artifacts, membership enforcement, exact/bounded/UNKNOWN routing, audit logs, documentation, security and privacy review, software and compute, workflow integration, user testing, and independent evaluation for one organization.","confidence":"LOW","assumptions":["One modeling language and one organizational integration are in scope.","No intervention, eligibility, enforcement, or other individual-level decision is automated.","A robust service requires multiple person-years across formal methods, software, testing, governance, and coordination.","May 2025 BLS annual mean wages include about $148,100 for software developers, $129,260 for mathematicians, and $99,730 for operations-research analysts before benefits and overhead."],"source_ids":["S7"]},"operational_launch":{"band_2026_usd":"250K_TO_1M","scope":"Independent assurance, a larger disclosed evaluation corpus, adversarial membership and label-propagation tests, software and compute hardening, accessibility and documentation work, compliance and records review, incident and rollback procedures, user training, coordination with policy-method owners, and a monitored model-analysis-only launch.","confidence":"LOW","assumptions":["A real adopter and accountable review owner have been identified before launch.","The restricted modes demonstrate useful coverage and performance.","UNKNOWN and timeout labels can be traced through every downstream report.","Evaluation compares the router with the adopter's existing simulation and reporting process; no live behavioral intervention is part of launch."],"source_ids":["S5","S6","S7"]},"annual_recurring":{"band_2026_usd":"250K_TO_1M","scope":"One organization's recurring labor for formal-language and solver maintenance, model-change review, software and compute, security and compliance upkeep, audit-log retention, periodic independent assurance, corpus refresh, user support, governance meetings, and evaluation of scope, label preservation, performance, and incidents.","confidence":"LOW","assumptions":["Approximately one to three full-time-equivalent staff plus fractional independent review and behavioral-policy governance are required.","Only one principal language family and a modest submission volume are maintained.","Major new reduction programs, regulated individual decisions, sensitive-data acquisition, and live intervention evaluation are excluded.","BLS wages are direct-labor anchors; the band additionally includes benefits, overhead, software, equipment, assurance, and coordination."],"source_ids":["S7"]}},"verified_pipeline_gates":{"externally_supported_problem":{"status":"NO","reason":"The bounded search found concern about prediction uncertainty and epistemic communication, but no evidence of an exact, terminating, unrestricted class-wide requirement. Primary and official sources instead describe uncertain model forecasts and empirical comparison.","source_ids":["S1","S5","S8"]},"externally_credible_adopter_or_authorizer":{"status":"UNCERTAIN","reason":"Behavioral-policy practitioners, policy makers, independent assessors, and governance personnel are credible role classes, but no named institution has accepted ownership or expressed demand for this workflow.","source_ids":["S5","S6","S7"]},"distinct_testable_incremental_claim":{"status":"YES","reason":"The remaining claim can be tested for one declared language: validate the reduction, enforce fragment membership, compare exact/bounded/UNKNOWN routing with binary timeout behavior, and test whether downstream artifacts preserve labels. Prior art makes the claim narrow but still falsifiable.","source_ids":["S3","S4","S7"]},"bounded_next_evidence_step":{"status":"YES","reason":"A time-limited artifact-and-interview audit can test whether the prerequisite universal requirement exists without implementing or deploying a policy system, using finite-horizon probabilistic requirements as the explicit comparator.","source_ids":["S1","S5","S8"]},"no_unresolved_safety_or_authority_stop":{"status":"YES","reason":"The next step is document research using public or voluntarily supplied organizational artifacts, makes no behavioral intervention or individual decision, and can stop before formal-tool development if demand is not demonstrated.","source_ids":["S6","S7"]},"credible_cost_scope_and_range":{"status":"YES","reason":"The bands explicitly include specialized labor, data and artifact handling, governance/compliance, coordination, software, equipment/compute, and independent evaluation. Current BLS wage data provide a direct-labor anchor, while deployment bands remain low-confidence because no language or adopter is fixed.","source_ids":["S7"]}},"next_evidence_step":"Before attempting a new proof, run a six-week, no-deployment prevalence audit with at most three behavioral-policy organizations and 20 existing specifications, model cards, procurement documents, or analysis protocols. Pre-register a qualifying requirement as one that demands an exact yes/no answer, termination for every model in an unbounded enforceable language class, and eventual rather than fixed-horizon attainment. Compare qualifying artifacts with finite-horizon probabilistic or scenario-analysis requirements, and have each apparent hit confirmed by both its model owner and policy-method owner. Falsify the problem premise for this sample if no artifact qualifies after clarification; advance to language formalization only if at least one current, decision-relevant requirement qualifies and its owner agrees that timeout or simulation is presently treated as definitive.","blocking_evidence":["A current behavioral-policy artifact that actually requires an exact, terminating, unrestricted class-wide verdict rather than calibrated, finite-horizon, or scenario-based prediction.","Confirmation from a named model owner and accountable policy-method owner that the requirement is decision-relevant and that current timeout or heuristic output is treated as definitive.","A declared behavioral-model language with formal syntax and operational semantics.","A mechanically or independently checked computable reduction preserving target attainment, or evidence that the chosen language cannot support it.","An enforceable restricted-fragment membership checker with proved termination and sound guarantee labels.","Representative-corpus evidence that the restricted or bounded modes have useful coverage and that UNKNOWN survives downstream reporting.","An adopter-specific comparison showing improvement over existing uncertainty, validation, and reporting controls."],"research_disposition":"PROBLEM_PREVALENCE_STUDY","world_novelty_boundary":"This was a bounded English-language web search emphasizing behavioral-policy prediction, agent-based social simulation, cognitive-model verification, ACT-R, neural-agent reachability, and official uncertainty/governance guidance. It found substantial technical collision: unbounded agent reachability proved undecidable by halting reduction with an implemented bounded fallback, plus established automated verification of cognitive models. It did not find the complete behavioral-policy-specific language router with audited exact/bounded/UNKNOWN propagation, but absence from this search is not evidence of world novelty, patentability, or freedom to operate."}