{"schema_version":1,"research_id":"eoa_inverse_innovation_exp05_external_evaluation_20260803","source_assessment_id":"computability_boundary_mapping__veterinary_medicine:P2:v0","cell_id":"computability_boundary_mapping__veterinary_medicine","search_queries":["site:nc3rs.org.uk computational modelling animal behaviour simulation replacement official","computational ethology animal behavior models reproducibility simulation fitting review","universal probabilistic programming halting inference undecidable trace matching","dovetailing semidecision halting problem executable simulator trace matching","site:braininitiative.nih.gov computational behavior tools animal official funding reproducibility","site:nc3rs.org.uk simulation modelling animal behaviour computational models official","site:olaw.nih.gov Guide Care Use Laboratory Animals IACUC veterinarian responsibilities official","site:nist.gov container security untrusted code sandbox guide","site:braininitiative.nih.gov animal behavior computational models funding opportunity","site:nih.gov computational ethology animal behavior funding modeling","site:nsf.gov computational ethology animal behavior funding program","site:simonsfoundation.org computational ethology animal behavior initiative","simulation-based inference official documentation simulator parameter posterior sbi toolkit","likelihood-free inference executable simulator parameter fitting animal behavior model","probabilistic programming nontermination inference universal language trace semantics research","exact trace matching simulator parameter search semi-decidable","\"Toward a science of computational ethology\" publication date","NIST SP 800-190 publication date","Open Logic Project Computability Theory 2026 revision","Brain Behavior Quantification Synchronization program publication date NIH"],"sources":[{"source_id":"S1","title":"Toward a Science of Computational Ethology","publisher":"Neuron, indexed by PubMed","url":"https://pubmed.ncbi.nlm.nih.gov/25277452/","source_class":"AUTHORITATIVE_SECONDARY","publication_date":"2014-10-01","accessed_at":"2026-08-03","claims_supported":["Computational ethology is an identifiable field enabled by automated measurement and analysis of animal behavior.","The field supplies a plausible setting for executable and computational behavior models, but this source does not document universal exact fitting services or timeout-as-no failures."]},{"source_id":"S2","title":"Brain Behavior Quantification and Synchronization Program","publisher":"NIH BRAIN Initiative","url":"https://www.braininitiative.nih.gov/research/systems-neuroscience/brain-behavior-quantification-and-synchronization-program","source_class":"GOVERNMENT_OR_REGULATOR","publication_date":"n.d.","accessed_at":"2026-08-03","claims_supported":["NIH is an identifiable funder seeking next-generation tools, analytic approaches, and computational models for complex behavior.","The program seeks mechanistic brain-behavior understanding and model development, but does not express demand for witness-only simulator matching."]},{"source_id":"S3","title":"Behavioral Systems","publisher":"U.S. National Science Foundation","url":"https://www.nsf.gov/funding/opportunities/behavioral-systems","source_class":"GOVERNMENT_OR_REGULATOR","publication_date":"n.d.","accessed_at":"2026-08-03","claims_supported":["NSF is an identifiable funder of animal-behavior research, including modeling and theoretical approaches.","The opportunity establishes an adjacent funding pathway but not an adoption commitment for the proposed protocol."]},{"source_id":"S4","title":"sbi: A toolkit for simulation-based inference","publisher":"Journal of Open Source Software","url":"https://joss.theoj.org/papers/10.21105/joss.02505.pdf","source_class":"PRIMARY_RESEARCH","publication_date":"2020-08-21","accessed_at":"2026-08-03","claims_supported":["Simulator-parameter fitting to match empirical observations is a documented scientific challenge.","The sbi toolkit performs Bayesian simulation-based inference for black-box stochastic simulators and is close functional prior art.","sbi estimates posterior regions and uncertainty rather than deciding exact existential trace matching for every unrestricted executable simulator."]},{"source_id":"S5","title":"A lambda-calculus foundation for universal probabilistic programming","publisher":"Microsoft Research / ACM ICFP","url":"https://www.microsoft.com/en-us/research/publication/a-lambda-calculus-foundation-for-universal-probabilistic-programming/","source_class":"PRIMARY_RESEARCH","publication_date":"2016-08","accessed_at":"2026-08-03","claims_supported":["Universal probabilistic-programming research already formalizes higher-order execution, random-sample traces, and trace-based inference.","This is adjacent prior art for executable stochastic simulators and replayable traces, although it does not provide the proposed witness-only veterinary workflow."]},{"source_id":"S6","title":"Computability Theory","publisher":"Open Logic Project","url":"https://builds.openlogicproject.org/content/computability/computability-theory/computability-theory.pdf","source_class":"AUTHORITATIVE_SECONDARY","publication_date":"2026-07-12","accessed_at":"2026-08-03","claims_supported":["The halting problem is undecidable for a universal partial-computation model.","Computably enumerable sets are semi-decidable: membership can eventually yield yes while nonmembership need never yield no.","Existential properties with a computable witness relation are computably enumerable.","Many-one reductions transfer computability status in the required source-to-target direction, and the text gives the standard construction of a machine that ignores its input and simulates a selected source computation."]},{"source_id":"S7","title":"Public Health Service Policy on Humane Care and Use of Laboratory Animals","publisher":"National Institutes of Health, Office of Laboratory Animal Welfare","url":"https://grants.nih.gov/policy-and-compliance/policy-topics/animal-welfare/laws-regulations/phs-policy","source_class":"GOVERNMENT_OR_REGULATOR","publication_date":"2015","accessed_at":"2026-08-03","claims_supported":["For covered live-vertebrate-animal activities, institutional assurances, IACUC approval, and veterinary authority are required.","IACUCs may approve, require changes, withhold approval, or suspend covered animal activities.","A software-only synthetic prototype does not itself involve live vertebrate animals, while later animal interventions would require the applicable institutional and regulatory review."]},{"source_id":"S8","title":"Application Container Security Guide, NIST SP 800-190","publisher":"National Institute of Standards and Technology","url":"https://nvlpubs.nist.gov/nistpubs/specialpublications/nist.sp.800-190.pdf","source_class":"STANDARD","publication_date":"2017-09","accessed_at":"2026-08-03","claims_supported":["Containers provide isolation mechanisms but retain risks including container escape, unbounded network access, privileged execution, unsafe system calls, and sensitive host mounts.","Executing arbitrary submitted simulator code therefore requires a specified and independently tested isolation architecture; ordinary containerization alone is not a complete safety argument."]}],"problem_evidence":{"support":"WEAK","rationale":"External sources establish that computational animal-behavior modeling exists and that fitting black-box simulator parameters to observations is difficult. They do not show that veterinary or computational-ethology laboratories currently promise a universal terminating exact-match service, report exhausted budgets as proof of no fit, or experience consequential decisions from that error. The mathematically coherent failure mode is therefore visible as a possibility, but its real-world prevalence and importance remain unverified.","source_ids":["S1","S4","S6"]},"stakeholder_evidence":{"support":"MODERATE","rationale":"NIH BRAIN and NSF are identifiable funders explicitly supporting animal-behavior tools and computational models, and a research PI could authorize an offline prototype. For later covered live-animal work, the IACUC and veterinarian are identifiable authorizers. None of these sources requests the particular witness-only protocol or demonstrates willingness to adopt it.","source_ids":["S2","S3","S7"]},"prior_art":{"proximity":"ADJACENT_PRIOR_ART","closest_analogues":[{"name":"sbi simulation-based inference toolkit","similarity":"Fits black-box stochastic simulator parameters to empirical observations and quantifies parameter uncertainty.","remaining_difference":"It performs approximate Bayesian inference over a declared prior and simulator workflow; it does not claim a total exact existential decider, complete positive dovetailing over arbitrary encoded programs, replay certificates, or bound-qualified negative labels.","source_ids":["S4"]},{"name":"Universal probabilistic programming with trace MCMC","similarity":"Supplies formal semantics for higher-order probabilistic programs and treats executions as functions of random-sample traces, with correctness conditions for trace-based inference.","remaining_difference":"It addresses probabilistic semantics and MCMC convergence, not a governed FOUND-WITNESS/UNKNOWN/NO-WITHIN-BOUND service for animal-behavior simulators.","source_ids":["S5"]},{"name":"Computably enumerable witness search and halting reductions","similarity":"The established theory already supplies the proposal's mathematical core: undecidable halting, semi-decidable existential properties, source-to-target reductions, and machines that ignore an input while simulating another computation.","remaining_difference":"The remaining contribution is an applied protocol combining fair scheduling, replayable simulator certificates, explicit operational budgets, bound-preserving labels, security controls, and veterinary research governance.","source_ids":["S6"]}],"distinctive_claim_remaining":"For an enforceably encoded class of partial animal-behavior simulators, a laboratory service can outperform conventional fitting specifically in guarantee honesty—not universal practical yield—by returning only independently replayable exact-match witnesses, UNKNOWN at an unrestricted operational budget, and NO-WITHIN-BOUND after complete finite-box exhaustion, while a fair ideal schedule prevents one looping candidate from starving every later finite witness. This is falsified if the candidate encoding or match predicate is not decidable, the schedule is unfair, a valid finite witness can be missed in the ideal schedule, replay checking disagrees with search output, or any budget-exhausted unrestricted query is labeled NO.","confidence":"MODERATE"},"implementation_evidence":{"support":"MODERATE","rationale":"The mathematical kernel is implementable when simulators, candidates, finite traces, and certificates have effective encodings: existential finite executions are semi-decidable, and a budgeted scheduler can always return UNKNOWN. A synthetic restricted-interpreter prototype is straightforward. Production feasibility is materially weaker because arbitrary code needs step-level control, deterministic replay contracts, resource accounting, and hardened isolation; NIST documents container-escape and configuration risks. No implementation, proof review, workflow integration, or user-label test has been observed.","source_ids":["S5","S6","S8"]},"scores":{"meaningful_impact":{"score":3,"rationale":"Preventing unsupported model rejection and overinterpretation could improve research decisions, but the frequency and downstream harm of the stated failure are not externally measured.","source_ids":["S1","S4"]},"stakeholder_pull":{"score":2,"rationale":"Funders visibly support computational behavior tools and models, but no laboratory, funder, or regulator expresses demand for this exact output contract.","source_ids":["S2","S3"]},"incremental_advantage":{"score":3,"rationale":"The protocol offers stronger semantic honesty and reproducible positive certificates than heuristic fitting, while generally sacrificing practical completeness and returning potentially many UNKNOWN results.","source_ids":["S4","S6"]},"distinctiveness_plausibility":{"score":2,"rationale":"The domain packaging may be distinctive, but its mathematical mechanisms are established computability practice and adjacent trace-based simulator inference already exists.","source_ids":["S4","S5","S6"]},"technical_implementability":{"score":4,"rationale":"A toy-language prototype with finite traces, round-robin step scheduling, certificates, and explicit labels is readily implementable; accepting arbitrary native simulator code safely is substantially harder.","source_ids":["S6","S8"]},"adoption_authority_feasibility":{"score":3,"rationale":"A PI can authorize research-only synthetic evaluation, while IACUC and veterinary authority are clearly available for later covered animal work. Actual adopter interest and local software-security approval remain unknown.","source_ids":["S2","S3","S7","S8"]},"evidence_readiness":{"score":4,"rationale":"Core scheduler, replay, labeling, bounded-exhaustion, and reduction obligations admit a small predeclared test suite and independent proof review, although prevalence and adoption require fieldwork.","source_ids":["S6"]},"safety_net_benefit":{"score":3,"rationale":"Explicit UNKNOWN and prohibition on biological or treatment interpretation reduce epistemic and animal-welfare risk, but benefit is prospective because no documented harmful deployment was found.","source_ids":["S6","S7"]},"scalability":{"score":2,"rationale":"Fair dovetailing is theoretically complete on positive instances but may allocate negligible practical work to deep candidates, produce mostly UNKNOWN, and multiply isolation and compute overhead across untrusted programs.","source_ids":["S6","S8"]}},"score_confidence":"MODERATE","costs":{"first_evidence":{"band_2026_usd":"50K_TO_250K","scope":"A bounded six-laboratory workflow/log study, independent reduction review, and restricted-interpreter prototype tested on at most twenty sealed synthetic simulators against serial, per-run-timeout, and finite-grid comparators.","confidence":"LOW","assumptions":["Approximately 0.5-1.0 research-software-engineer FTE for three to six months.","Part-time computability reviewer, computational-ethology investigator, security reviewer, and study coordination.","No animals, clinical records, proprietary production simulators, or production cloud service.","Resource-equivalent estimate rather than a vendor quote or market-price measurement."],"source_ids":["S4","S6","S8"]},"initial_deployment_startup":{"band_2026_usd":"250K_TO_1M","scope":"Build a production-oriented simulator representation, step-controlled runtime, fair scheduler, certificate checker, result schema, audit records, sandbox architecture, and one institutional integration.","confidence":"LOW","assumptions":["Two to five engineering/research FTE-equivalents for six to twelve months.","Only one or a small number of explicitly supported simulator languages.","Independent proof, security, and scientific-workflow reviews are included.","No live-animal intervention module or clinical decision support."],"source_ids":["S5","S6","S8"]},"operational_launch":{"band_2026_usd":"250K_TO_1M","scope":"Security hardening, penetration and escape testing, monitoring, documentation, training, and a controlled multi-laboratory research launch with rollback capability.","confidence":"LOW","assumptions":["Pilot deployment at two to five institutions.","Institutional security and research-governance review is required.","Untrusted native code is excluded unless the isolation architecture passes independent testing.","Any live-animal use is separately reviewed under applicable IACUC and veterinary authority."],"source_ids":["S7","S8"]},"annual_recurring":{"band_2026_usd":"250K_TO_1M","scope":"Operate a small multi-institution research service, including compute, storage, security patching, simulator-runtime maintenance, certificate compatibility, user support, audits, and periodic guarantee re-review.","confidence":"LOW","assumptions":["One to four engineering/operations FTE-equivalents plus scientific and security review.","Compute demand remains research-scale; pathological submissions are quota-limited.","No estimate of actual user volume, simulator depth, or cloud pricing was obtained.","Costs could fall below this band for a single-lab offline tool or exceed it for unrestricted public submissions."],"source_ids":["S2","S6","S8"]}},"verified_pipeline_gates":{"externally_supported_problem":{"status":"UNCERTAIN","reason":"Simulator fitting and computational animal-behavior modeling are externally supported, but the defining universal-Boolean requirement, timeout-as-no practice, prevalence, and consequence were not documented.","source_ids":["S1","S4"]},"externally_credible_adopter_or_authorizer":{"status":"YES","reason":"NIH BRAIN and NSF are credible adjacent funders; a research PI can authorize an offline prototype; and IACUC/veterinary authorities are externally specified for covered animal activities. Solution-specific adoption remains unproven.","source_ids":["S2","S3","S7"]},"distinct_testable_incremental_claim":{"status":"YES","reason":"Fair witness discovery, replay accuracy, output labeling, and finite-bound negatives can be compared directly with serial search, serial timeouts, and finite-grid search under predeclared cases and falsifiers.","source_ids":["S4","S6"]},"bounded_next_evidence_step":{"status":"YES","reason":"A small partnered workflow study plus a sealed twenty-simulator suite and independent proof review can be completed with fixed sites, cases, comparators, budgets, and pass/fail rules.","source_ids":["S4","S6"]},"no_unresolved_safety_or_authority_stop":{"status":"UNCERTAIN","reason":"Synthetic restricted-language testing can avoid animals and unsafe native execution, but production acceptance of arbitrary executable simulators has unresolved isolation risks, and jurisdiction-specific authority beyond U.S. PHS-covered animal work was not assessed.","source_ids":["S7","S8"]},"credible_cost_scope_and_range":{"status":"UNCERTAIN","reason":"The four bands have explicit staffing, deployment, and security scope, but no compensation benchmarks, infrastructure quotes, workload measurements, or actual institutional estimates were obtained.","source_ids":["S2","S8"]}},"next_evidence_step":"Recruit six computational-ethology or veterinary-behavior modeling laboratories for a predeclared workflow study. Audit their latest thirty simulator-fitting jobs for unrestricted Boolean language, simulator nontermination, individual or global timeout handling, and whether budget exhaustion became a negative model conclusion. In parallel, give each lab the restricted prototype and a sealed suite of at most twenty toy simulators containing immediate witnesses, long-running witnesses, early infinite loops, malformed encodings, and known-empty finite boxes. Compare (A) serial execution without per-run timeout, (B) serial execution with a fixed per-run timeout, (C) finite-grid exhaustion, and (D) fair step-dovetailing with FOUND-WITNESS/UNKNOWN/NO-WITHIN-BOUND. Require independent replay of every witness and a second reviewer to check the halting-to-matching reduction. Falsify the opportunity if none of the audited workflows makes the claimed universal or timeout-to-no error; falsify the intervention if D starves any finite witness in the ideal schedule, emits an unreplayable witness, emits unrestricted NO at budget exhaustion, loses finite bounds, or users reliably interpret FOUND-WITNESS as biological validation despite training.","blocking_evidence":["No external source documents an actual veterinary or computational-ethology laboratory promising universal exact Boolean simulator matching.","No prevalence estimate or workflow log shows that timeouts are converted to no-fit conclusions.","No named laboratory has committed to adopt or fund the witness-only protocol.","The exact source-to-target reduction and representation-preservation obligations have not been independently checked.","Fair step-level scheduling and deterministic replay have not been demonstrated for real simulator languages.","No independently tested sandbox design establishes acceptable containment of arbitrary submitted simulator code.","User comprehension of FOUND-WITNESS, UNKNOWN, and NO-WITHIN-BOUND has not been tested.","Cost bands lack compensation benchmarks, infrastructure quotes, utilization measurements, and institutional estimates.","Legal and animal-welfare authority outside U.S. PHS-covered research remains unassessed."],"research_disposition":"PROBLEM_PREVALENCE_STUDY","world_novelty_boundary":"This eight-source evaluation found established computability theory, universal probabilistic-program trace semantics, black-box simulation-based inference, and adjacent animal-behavior funding programs. It did not find a direct veterinary witness-only matching service, but the search was neither exhaustive nor a world-novelty review. World novelty, patentability, freedom to operate, market size, realized impact, and comprehensive prior-art coverage remain unmeasured.","arm":"COMPLETE_PROPOSAL_PORTFOLIO","candidate_version":0,"controller_recommendation":{"action":"STOP_EMPIRICAL_RESEARCH_NEEDED","repairable":false,"material_progress_observed":false,"progress_targets":["Obtain workflow logs and structured interviews from at least six target laboratories to measure the prevalence and consequence of universal or timeout-as-no claims.","Secure at least one named laboratory partner willing to test the result contract in a research-only workflow.","Have an independent reviewer validate the exact simulator encoding, quantifiers, computable construction, reduction direction, and answer preservation.","Demonstrate the four comparators on the sealed twenty-simulator suite with replayable certificates and machine-checked result labels.","Specify and independently test the execution boundary for supported simulator languages, including step control, network denial, filesystem isolation, quotas, and rollback.","Test whether target users distinguish existence witnesses, unknown results, finite-bound negatives, and biological model validation.","Replace resource-equivalent cost assumptions with staffing rates, infrastructure quotes, and measured workload estimates."],"reason":"Bounded web research supports the mathematical classification, adjacent modeling need, and identifiable funders and animal-research authorizers, but it cannot establish that the stated laboratory failure visibly occurs, that users will accept UNKNOWN, or that the protocol works safely with real simulator runtimes. Those decisive uncertainties require workflow fieldwork, proprietary logs, user testing, and live prototype evaluation rather than further bounded web search."},"proposal_index":2}