{"schema_version":1,"research_id":"eoa_inverse_innovation_exp05_external_evaluation_20260803","source_assessment_id":"computability_boundary_mapping__veterinary_medicine:P1:v0","cell_id":"computability_boundary_mapping__veterinary_medicine","search_queries":["veterinary clinical decision support software medication safety programmable rules policy official veterinary informatics","model checking clinical guidelines executable medical protocols formal verification PROforma","veterinary telemedicine AI clinical decision support veterinarian authority official policy","formal verification medical device software finite state model checking safety standard IEC 62304","site:cvo.org artificial intelligence veterinary medicine guidance veterinarian 2026 PDF","site:fda.gov animal veterinary artificial intelligence software veterinarian clinical decision support","Rice theorem original paper 1953 classes recursively enumerable sets pdf","BLS Occupational Employment Wage Statistics software developers veterinarians May 2025","site:avma.org artificial intelligence veterinary clinical decision support policy veterinarian accountability","site:aaha.org veterinary artificial intelligence guidelines clinical decision support risks","site:avma.org veterinary technology AI policy 2025 veterinarian decision","veterinary clinical decision support system study medication errors electronic prescribing animals","site:plato.stanford.edu computability halting problem Rice theorem program properties","site:cs.cornell.edu Rice theorem undecidable semantic properties programs lecture notes","site:cambridge.org Rice theorem program properties undecidable","site:openlogicproject.org Rice theorem computability","\"The Syntax and Semantics of the PROforma Guideline Modeling Language\" DOI","\"Model Checking for Clinical Guidelines: an Agent-based Approach\" DOI","\"Error reporting in a large animal veterinary teaching hospital\" DOI full text","site:europepmc.org PMC212780 PROforma","FDA \"Veterinarian-Client-Patient Relationships, Prescribing/Dispensing Animal Drugs and Telemedicine\" date","Frontiers Medical Errors Cause Harm in Veterinary Hospitals publication date 2019","PMC1839493 \"Model Checking for Clinical Guidelines\" publication 2006","PMC6370638 publication date"],"sources":[{"source_id":"S1","title":"Medical Errors Cause Harm in Veterinary Hospitals","publisher":"Frontiers in Veterinary Science","url":"https://www.frontiersin.org/journals/veterinary-science/articles/10.3389/fvets.2019.00012/full","source_class":"PRIMARY_RESEARCH","publication_date":"2019-02-05","accessed_at":"2026-08-03","claims_supported":["A retrospective study examined 560 incident reports from three veterinary hospitals.","Drug errors were the most frequently reported error type, and 15% of reported incidents resulted in patient harm.","The authors caution that voluntary reporting underestimates occurrence and does not establish true prevalence.","The study supports the importance of incident review and veterinary patient-safety interventions."]},{"source_id":"S2","title":"Use of Medical Devices Enabled by Artificial Intelligence in the Practice of Veterinary Medicine","publisher":"College of Veterinarians of Ontario","url":"https://www.cvo.org/standards/use-of-medical-devices-enabled-by-artificial-intelligence-in-the-practice-of-veterinary-medicine","source_class":"OFFICIAL_GUIDANCE","publication_date":"2024-12","accessed_at":"2026-08-03","claims_supported":["The veterinary regulator identifies a need for information enabling veterinarians to assess technology risk and potential harm.","The guidance calls for transparency about purpose, methods, data, testing, performance, monitoring, risks, and benefits of decision-support tools.","Veterinarians remain accountable for their choices and must use clinical reasoning rather than defer to automated recommendations.","The guidance reports limited regulatory oversight for most animal-health medical devices and describes irresponsible use as posing significant risk of harm."]},{"source_id":"S3","title":"Veterinarian-Client-Patient Relationships, Prescribing/Dispensing Animal Drugs and Telemedicine","publisher":"U.S. Food and Drug Administration, Center for Veterinary Medicine","url":"https://www.fda.gov/animal-veterinary/product-safety-information/veterinarian-client-patient-relationships-prescribingdispensing-animal-drugs-and-telemedicine","source_class":"GOVERNMENT_OR_REGULATOR","publication_date":"2024","accessed_at":"2026-08-03","claims_supported":["Prescription animal drugs may be dispensed only by or on the lawful order of a licensed veterinarian acting professionally.","Licensed-veterinarian involvement and oversight are required for safe prescription-drug use under federal requirements.","Extralabel prescribing requires a valid veterinarian-client-patient relationship, while state and local requirements may add obligations.","An automated verifier cannot replace the veterinarian's legally required prescribing authority."]},{"source_id":"S4","title":"Recursive Functions","publisher":"Stanford Encyclopedia of Philosophy, Stanford University","url":"https://plato.stanford.edu/archives/spr2026/entries/recursive-functions/","source_class":"AUTHORITATIVE_SECONDARY","publication_date":"2024-03-01","accessed_at":"2026-08-03","claims_supported":["Rice's theorem states that every nontrivial index set representing a semantic property of partial computable functions is undecidable.","The entry identifies program totality and correctness with respect to a specified function as examples of undecidable decision problems.","It explains the required many-one reduction direction: reduce a known noncomputable set to the target problem.","The theorem applies only after the policy representation and property are shown to match its assumptions."]},{"source_id":"S5","title":"The Syntax and Semantics of the PROforma Guideline Modeling Language","publisher":"Journal of the American Medical Informatics Association / American Medical Informatics Association","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC212780/","source_class":"PRIMARY_RESEARCH","publication_date":"2003","accessed_at":"2026-08-03","claims_supported":["PROforma is an executable process-modeling language used for clinical decision support, guidelines, and other clinical applications.","The work defines public syntax, operational semantics, expression semantics, and task-state operations for executable guidelines.","The language includes decisions, plans, actions, data queries, and externally initiated operations.","Executable clinical-policy languages with formal semantics substantially predate this veterinary proposal."]},{"source_id":"S6","title":"Model Checking for Clinical Guidelines: an Agent-based Approach","publisher":"American Medical Informatics Association Annual Symposium Proceedings","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC1839493/","source_class":"PRIMARY_RESEARCH","publication_date":"2006","accessed_at":"2026-08-03","claims_supported":["The authors translated executable clinical guidelines and patient/environment behavior into Promela and checked LTL properties with SPIN.","SPIN converted processes into finite automata and produced counterexample paths when properties failed.","The proof-of-concept found inconsistencies in a stroke guideline not previously detected by hand.","The work demonstrates feasibility but reports ongoing experimentation and human-medical rather than veterinary deployment."]},{"source_id":"S7","title":"IEC 62304:2006 Medical device software — Software life cycle processes","publisher":"International Organization for Standardization / International Electrotechnical Commission","url":"https://www.iso.org/cms/live/live/en/sites/isoorg/contents/data/standard/03/84/38421.html?browse=tc","source_class":"STANDARD","publication_date":"2006-05","accessed_at":"2026-08-03","claims_supported":["IEC 62304 defines life-cycle requirements and a common process framework for medical-device software.","The standard remains current following confirmation in 2021, although it is under systematic review.","It supports disciplined development and change control but does not itself establish clinical correctness, computability, or veterinary regulatory applicability."]},{"source_id":"S8","title":"Occupational Employment and Wages — May 2025","publisher":"U.S. Bureau of Labor Statistics","url":"https://www.bls.gov/news.release/ocwage.htm","source_class":"GOVERNMENT_OR_REGULATOR","publication_date":"2026-05-15","accessed_at":"2026-08-03","claims_supported":["May 2025 mean annual wages were $148,100 for software developers, $153,930 for computer and information research scientists, and $142,680 for veterinarians.","The corresponding mean hourly wages were $71.20, $74.00, and $68.60.","These wage figures provide a resource-equivalent labor floor; benefits, overhead, specialized consulting premiums, infrastructure, validation, and procurement require additional assumptions."]}],"problem_evidence":{"support":"MODERATE","rationale":"Veterinary patient-safety harm and frequent drug errors are directly documented, and veterinary regulatory guidance identifies risks, transparency needs, and weak oversight for advanced decision-support tools. However, no searched source verifies the proposal's defining premise: an actual multispecies platform that accepts arbitrary executable care policies and promises a total exact Boolean verifier over every policy and trajectory. The general safety problem exists, but the stated computability-boundary failure remains an unverified platform-specific hypothesis.","source_ids":["S1","S2","S3","S4"]},"stakeholder_evidence":{"support":"MODERATE","rationale":"The College of Veterinarians of Ontario is an identifiable professional regulator expressing a need for transparent risk, validation, and monitoring information, while FDA requirements identify the licensed veterinarian and relevant veterinary boards as prescribing authorities. These sources support authorizer interest in safe decision support, but no veterinary hospital, platform owner, funder, or safety committee is documented as requesting this particular verifier or committing resources to adopt it.","source_ids":["S2","S3"]},"prior_art":{"proximity":"SUBSTANTIAL_COLLISION","closest_analogues":[{"name":"PROforma executable clinical-guideline language","similarity":"Provides formal syntax and operational semantics for executable clinical decisions, plans, actions, data, and task-state transitions.","remaining_difference":"The reviewed source is human-medical and does not establish a veterinary-specific finite fragment, computability decision record, or abstention router.","source_ids":["S5"]},{"name":"GLARE/SPIN model checking of clinical guidelines","similarity":"Translates executable guidelines plus patient and environment behavior into finite automata, checks temporal properties, and returns counterexample paths.","remaining_difference":"The proof-of-concept does not address the proposed veterinary policy language, prohibited-dose table, explicit UNKNOWN/OUT-OF-SCOPE workflow, or a checked reduction for an unrestricted language.","source_ids":["S6"]},{"name":"Rice-theorem boundary analysis","similarity":"Directly supports rejecting universal exact semantic analyses of unrestricted program classes when theorem assumptions and reduction obligations are met.","remaining_difference":"No source supplies the required reduction from the actual veterinary language; the candidate's platform, semantics, and universal claim have not been obtained.","source_ids":["S4"]},{"name":"IEC 62304 software life-cycle framework","similarity":"Provides established safety-oriented software-development and maintenance processes relevant by analogy to clinical software.","remaining_difference":"It is not a computability classification method and its direct applicability to veterinary decision-support software is not established.","source_ids":["S7"]}],"distinctive_claim_remaining":"For a named veterinary platform whose current unrestricted policy language and universal verifier promise are documented, an enforceable finite-state fragment plus conservative transition abstraction and labeled abstention router will produce zero false VERIFIED-WITHIN-MODEL results against complete bounded enumeration, prevent every out-of-fragment or resource-exhausted case from reaching execution, and retain useful coverage compared with simulation-only review; separately, an independently checked source-to-target reduction will establish undecidability only for the exactly matched unrestricted language and property. Each clause is falsified respectively by one covered false clearance, one fragment or routing bypass, predeclared inadequate coverage, or failure of the reduction's totality, computability, answer preservation, or scope match.","confidence":"HIGH"},"implementation_evidence":{"support":"MODERATE","rationale":"Formal executable guideline languages and SPIN-based model checking demonstrate that a toy finite-state implementation is technically credible. Finite admission rules, exhaustive enumeration, counterexample traces, and explicit abstention are conventional software artifacts. The main unresolved implementation obligations are veterinary rather than purely computational: proving the patient-state abstraction conservative for its declared scope, maintaining multispecies contraindication data, integrating external-service semantics, preventing runtime bypass, controlling alarm burden, documenting changes, and securing veterinarian authorization. IEC 62304 supplies a lifecycle analogue but not a veterinary clinical-validation shortcut.","source_ids":["S2","S3","S5","S6","S7"]},"scores":{"meaningful_impact":{"score":4,"rationale":"Medication and other clinical errors visibly harm veterinary patients, so preventing a false software clearance could matter substantially. Realized benefit is unmeasured and depends on the hypothesized platform existing.","source_ids":["S1","S2"]},"stakeholder_pull":{"score":2,"rationale":"Regulatory guidance expresses demand for transparency, risk assessment, testing, and professional control, but no stakeholder explicitly requests, funds, or adopts the proposed computability-boundary system.","source_ids":["S2","S3"]},"incremental_advantage":{"score":2,"rationale":"Explicit abstention and enforceable scope could improve on simulation plus Boolean timeout handling, but executable guideline semantics and model checking are already demonstrated; superiority in veterinary workflow has not been tested.","source_ids":["S5","S6"]},"distinctiveness_plausibility":{"score":2,"rationale":"The veterinary adaptation and combined governance record may be distinctive, but the technical core substantially collides with long-established computability theory, executable clinical-guideline languages, and clinical model checking.","source_ids":["S4","S5","S6"]},"technical_implementability":{"score":3,"rationale":"The bounded synthetic demonstrator is implementable with established tools. A sound, clinically relevant multispecies abstraction and bypass-proof production integration remain difficult and unverified.","source_ids":["S5","S6","S7"]},"adoption_authority_feasibility":{"score":3,"rationale":"Licensed veterinarians and veterinary regulators provide a recognizable authority chain, and an offline research step preserves that authority. Multi-jurisdiction deployment and software oversight remain unsettled.","source_ids":["S2","S3"]},"evidence_readiness":{"score":2,"rationale":"The candidate supplies a precise test concept, but no real platform specification, policies, trajectory model, adopter commitment, contraindication corpus, proof artifact, or prototype result is externally available.","source_ids":["S2","S4","S6"]},"safety_net_benefit":{"score":4,"rationale":"Fail-closed fragment admission, distinct UNKNOWN and OUT-OF-SCOPE states, counterexample traces, and veterinarian review could materially reduce deceptive clearance if correctly enforced.","source_ids":["S2","S3","S6"]},"scalability":{"score":2,"rationale":"The software machinery can be reused, but every species, clinical domain, external service, and language change may require new semantics, safety properties, abstraction validation, and review; state-space growth is unresolved.","source_ids":["S5","S6","S7"]}},"score_confidence":"MODERATE","costs":{"first_evidence":{"band_2026_usd":"10K_TO_50K","scope":"One 4- to 8-week offline study defining the toy grammar, exhaustive oracle, 40-100 synthetic policies, model-checker translation, router tests, reduction draft, and independent review; excludes live animal data and production integration.","confidence":"MODERATE","assumptions":["Approximately 120-240 formal-methods/software hours and 24-60 veterinarian-review hours.","Existing open-source model-checking infrastructure is used.","BLS mean hourly wages are treated as salary-equivalent floors and multiplied by roughly 1.5-2.0 for benefits, overhead, and specialist premiums.","No patent, regulatory submission, or clinical-effect study is included."],"source_ids":["S6","S8"]},"initial_deployment_startup":{"band_2026_usd":"250K_TO_1M","scope":"Productize the fragment compiler and verifier, integrate one platform in a sandbox, create traceability and change controls, curate one narrow veterinary safety-property set, complete security and independent formal review, and run nonclinical acceptance testing.","confidence":"LOW","assumptions":["Roughly 1-3 engineer-years plus 0.25-0.75 veterinarian-year and specialized formal-methods review.","One platform, one jurisdiction, and one narrow clinical domain are covered.","No live efficacy trial or regulator-mandated premarket submission is assumed.","Integration complexity and proprietary data access are unknown."],"source_ids":["S2","S3","S7","S8"]},"operational_launch":{"band_2026_usd":"1M_TO_5M","scope":"Launch in a limited veterinary network with production integration, multispecies safety-data governance, validation, monitoring, incident response, training, legal review, staged rollout, and rollback capability.","confidence":"LOW","assumptions":["A named platform partner and veterinary network participate.","Multiple engineers, clinical informaticians, veterinarians, security personnel, and independent reviewers contribute for 12-24 months.","Costs exclude a large prospective clinical trial, damages, insurance changes, and broad international regulatory clearance.","Clinical-model development and integration, not model-checker compute, dominate cost."],"source_ids":["S1","S2","S3","S7","S8"]},"annual_recurring":{"band_2026_usd":"250K_TO_1M","scope":"Maintain two to five resource-equivalent roles for policy-language governance, contraindication updates, re-verification, incident review, security, user support, and independent periodic review for a limited deployment.","confidence":"LOW","assumptions":["BLS mean annual wages of roughly $143K-$154K for relevant occupations are salary floors before benefits and overhead.","Major language, platform, or jurisdiction expansion is treated as new startup work rather than routine maintenance.","Cloud model-checking expense is secondary to specialist labor.","The verifier remains decision support and does not independently authorize treatment."],"source_ids":["S2","S3","S7","S8"]}},"verified_pipeline_gates":{"externally_supported_problem":{"status":"UNCERTAIN","reason":"Veterinary errors and decision-support risks are externally supported, but the essential claimed failure—a real unrestricted veterinary policy language paired with a universal exact verifier requirement—was not found.","source_ids":["S1","S2","S4"]},"externally_credible_adopter_or_authorizer":{"status":"YES","reason":"The College of Veterinarians of Ontario is an identifiable regulator expressing decision-support risk and transparency needs, and FDA requirements establish licensed veterinarians and veterinary boards as authorities over prescription use. No committed adopter is identified, but credible authorizers are.","source_ids":["S2","S3"]},"distinct_testable_incremental_claim":{"status":"YES","reason":"The remaining claim specifies enforceable fragment admission, zero false verified verdicts against exhaustive bounded ground truth, correct abstention routing, comparative coverage, and independently reviewable reduction obligations.","source_ids":["S4","S5","S6"]},"bounded_next_evidence_step":{"status":"YES","reason":"A synthetic, fully enumerable, nonclinical comparison can be completed in one short study with explicit comparators, metrics, and stopping rules.","source_ids":["S6","S8"]},"no_unresolved_safety_or_authority_stop":{"status":"YES","reason":"The authorized next step uses synthetic data, makes no treatment decision, cannot issue medication, and can preserve licensed-veterinarian authority. Production deployment would require fresh jurisdictional, clinical-model, privacy, and safety review.","source_ids":["S2","S3"]},"credible_cost_scope_and_range":{"status":"YES","reason":"All four estimates state scope and exclusions and are anchored to current official occupational wage data, though platform integration and clinical-validation uncertainty keep later-stage confidence low.","source_ids":["S7","S8"]}},"next_evidence_step":"With a named platform partner, run one 4- to 8-week offline preregistered test: freeze a toy grammar of at most eight control states, three three-valued observations, five decision steps, no recursion or dynamic calls, and a fixed prohibited-action table; generate 40-100 synthetic policies and exhaustively enumerate every trajectory to create an independent oracle. Compare (A) sampled scenario simulation with timeout coerced to Boolean, (B) complete bounded enumeration, and (C) the proposed admission checker, SPIN-style model checker, and four-state router. Primary metrics are false VERIFIED-WITHIN-MODEL count, fragment-bypass count, UNKNOWN/OUT-OF-SCOPE conversion count, counterexample-trace fidelity, coverage, run time, and veterinarian review time. Stop and reject the intervention if C produces any covered false verification, permits any rejected policy to execute, collapses any abstention into authorization, or disagrees with the oracle without a documented oracle error; reject incremental advantage if C does not improve false-clearance behavior over A or if a predeclared majority of representative in-scope policies abstain. In parallel, have an independent reviewer check the unrestricted reduction's source status, source-to-target direction, totality, computability, biconditional answer preservation, encoding, and exact quantifiers. This step cannot establish clinical effectiveness, production soundness, or unrestricted world-level solvability without access to the actual platform semantics.","blocking_evidence":["No external evidence that the hypothesized multispecies platform accepts arbitrary executable care policies or promises a universal terminating safe/unsafe verifier.","No named veterinary hospital, platform owner, funder, or safety committee has committed to the study or supplied platform artifacts.","No validated conservative patient-state abstraction or maintained multispecies prohibited-action table exists for the claimed guarantee.","No independently checked reduction matches an actual veterinary policy language and computation model.","No empirical comparison establishes useful coverage, manageable alarm burden, workflow latency, or safe escalation behavior.","No jurisdiction-specific determination covers software classification, privacy, professional liability, state-board requirements, and production authority.","Later-stage cost estimates lack platform integration quotes, data-curation estimates, and any regulator-required validation scope."],"research_disposition":"PARTNERED_RESEARCH_PROGRAM","world_novelty_boundary":"This eight-source evaluation found substantial adjacent and colliding prior art in executable clinical-guideline languages, clinical model checking, computability theory, and medical-software lifecycle practice. It did not measure world novelty, patentability, freedom to operate, market size, or realized impact, and it did not conduct patent, product-catalog, proprietary-platform, or exhaustive literature searches. Any novelty claim is therefore limited to the unverified veterinary-specific combination and remains unmeasured.","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 veterinary platform and clinical-safety partner with authority to disclose the accepted policy language, current verifier contract, representative policies, and workflow constraints.","Document whether the alleged unrestricted total-exact requirement actually exists; if the production class is already finite, reclassify the work as complexity and safety engineering rather than computability-boundary mapping.","Freeze executable semantics, observation and external-service models, prohibited-action property, quantifiers, and output meanings in a reviewable specification.","Complete the bounded comparator study with exhaustive ground truth and preregistered falsifiers, including fragment-bypass and abstention-laundering tests.","Produce and independently check the source-to-target reduction against the actual unrestricted language; record unresolved scope rather than inferring undecidability from timeouts.","Obtain a jurisdiction-specific authority and safety memorandum plus written adopter or funder intent for any work beyond the synthetic study.","Replace preliminary resource-equivalent estimates with partner staffing commitments, integration estimates, data-curation effort, and external-review quotes."],"reason":"Open-web evidence establishes an important adjacent veterinary safety problem, credible professional authority, substantial prior-art collision, and a feasible synthetic test. It does not establish the defining platform-specific problem or field performance. Resolving those gaps requires proprietary platform semantics, a participating adopter, independent proof review, and empirical offline testing; production claims would additionally require clinical and workflow evidence. Under the controller rule, that evidence cannot be completed by further bounded web search, so empirical research is required."},"proposal_index":1}