{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp04_retrieval_first_paired20_20260802","cell_id":"computability_boundary_mapping__public_administration_policy","hypothesis_id":"H5","search_queries":["public procurement AI assurance case vendor claims evidence independent review algorithmic systems","government procurement automated decision systems contract technical requirements limitations fallback monitoring","formal methods public procurement software verification proof obligations government contracts","AI procurement standard supplier claims intended purpose limitations change re-evaluation ISO NIST","\"computability\" procurement algorithm government","assurance case procurement software supplier claims formal verification standard"],"sources":[{"source_id":"S1","title":"Assurance case — Glossary","publisher":"National Institute of Standards and Technology","url":"https://csrc.nist.gov/glossary/term/assurance_case","source_class":"OFFICIAL_GUIDANCE","claims_supported":["NIST defines an assurance case as a reasoned, auditable artifact supporting explicit claims through systematic argumentation, evidence, and explicit assumptions.","NIST also identifies the established systems-assurance formulation of structured arguments and evidence for specified system properties."]},{"source_id":"S2","title":"Incorporating Software Requirements into the System RFP: Survey of RFP Language for Software by Topic, v. 2.0","publisher":"Carnegie Mellon University Software Engineering Institute","url":"https://www.sei.cmu.edu/documents/1864/2009_003_001_15017.pdf","source_class":"AUTHORITATIVE_SECONDARY","claims_supported":["A government-software RFP template already requires suppliers to submit an ISO/IEC 15026 assurance case containing claims, arguments, evidence, explicit assumptions, and the qualifications of an approving authority.","The template provides that the assurance case becomes part of the contract and serves as an initial acceptance condition, creating a close pre-award and contractual analogue to the nominated gate."]},{"source_id":"S3","title":"Structured Assurance Case Metamodel Specification Version 2.3","publisher":"Object Management Group","url":"https://www.omg.org/spec/SACM","source_class":"STANDARD","claims_supported":["SACM is an established standard for representing structured assurance cases.","The standard supplies a reusable representation for auditable claims, arguments, evidence, scope, context, and assumptions, including communication between suppliers, acquirers, operators, and regulators."]},{"source_id":"S4","title":"For software customers — Software Security Code of Practice","publisher":"UK National Cyber Security Centre","url":"https://www.ncsc.gov.uk/section/software-security-code-of-practice/software-customers","source_class":"OFFICIAL_GUIDANCE","claims_supported":["Software purchasers are instructed to request and assess a vendor Assurance Principles and Claims document and use it in supplier negotiations and contracts.","For higher assurance, purchasers are directed toward third-party independent testing and toward periodic updates and continuing verification of controls."]},{"source_id":"S5","title":"Guide on the Scope of the Directive on Automated Decision-Making","publisher":"Government of Canada","url":"https://www.canada.ca/en/government/system/digital-government/digital-government-innovations/responsible-use-ai/guide-scope-directive-automated-decision-making.html","source_class":"GOVERNMENT_OR_REGULATOR","claims_supported":["Canada's automated-decision requirements cover procured systems and require impact assessment, quality controls, transparency, recourse, and public reporting.","A significant modification to system scope or functionality triggers renewed treatment as a new system, providing an official change-trigger analogue.","The guide expressly includes fraud-probability estimation among covered administrative applications."]},{"source_id":"S6","title":"Public Procurement for Responsible AI? Understanding U.S. Cities’ Practices and Needs","publisher":"2nd Workshop on Regulatable ML at NeurIPS 2024","url":"https://openreview.net/attachment?id=pw6AEHJfvM&name=pdf","source_class":"PRIMARY_RESEARCH","claims_supported":["Interviews with 18 employees across seven U.S. cities found vendor opacity and trade-secrecy assertions obstructing assessment of procured AI.","Participants reported inadequate disclosure of performance information, occasional buyer-run independent evaluations, and concern that vendor demonstrations selectively show successful cases.","The study supports a general procurement-assurance problem but does not document acceptance of universal, exact, terminating detection claims."]}],"proximity":"SUBSTANTIAL_COLLISION","closest_analogues":[{"name":"Government RFP software-assurance-case submission","similarity":"It already places a claim-evidence-assumption record in the solicitation, evaluates it before award, identifies an approving authority, incorporates it into the contract, and uses it as an acceptance condition.","remaining_difference":"Its exemplified claims concern software safety and security generally; it does not mandate a model-relative computability classification, universal-quantifier audit, termination proof, or undecidability reduction.","source_ids":["S1","S2"]},{"name":"Structured Assurance Case Metamodel","similarity":"It standardizes auditable, scoped claims supported by arguments, evidence, context, and explicit assumptions for communication between suppliers and acquirers.","remaining_difference":"It is claim-domain neutral and supplies neither computability-specific proof obligations nor a procurement decision rule for rejecting impossible total-exact requirements.","source_ids":["S1","S3"]},{"name":"NCSC supplier-claims procurement assurance","similarity":"It instructs buyers to obtain vendor claim-and-evidence records, independently test higher-assurance claims, use findings contractually, and maintain assurance after purchase.","remaining_difference":"Its subject is cybersecurity resilience and control effectiveness rather than whether a declared problem class admits a correct terminating algorithm.","source_ids":["S4"]},{"name":"Canadian automated-decision governance and material-change trigger","similarity":"It applies to procured public algorithms, requires quality and impact controls, and reopens governance obligations after material scope or functionality changes.","remaining_difference":"It assesses operational risk and administrative-law safeguards, not the computability status of vendor capability guarantees.","source_ids":["S5"]}],"overlapping_components":["Pre-award technical assurance gate","Explicit top-level claims and scope","Evidence and argument requirements","Assumption register","Independent or qualified review","Contract incorporation and acceptance conditions","Ongoing monitoring and assurance maintenance","Material-change re-evaluation trigger","Public-sector automated-decision applicability"],"remaining_contrastive_claim":"Within public algorithm procurement, the still-distinct element is a mandatory computability-specific subcase that formalizes the input class and universal quantifiers and requires model-matched evidence of total decidability—or an appropriately scoped impossibility result—before an exact terminating guarantee can be awarded or renewed.","claim_falsifier":"The contrastive claim would be falsified by an earlier public-procurement standard, operative RFP template, procurement program, product, or patent that requires pre-award model-relative decidability or totality classification, checks constructive proofs or reductions, binds the resulting guarantee contractually, and reopens it after material system changes.","problem_support":"MODERATE","recommendation":"RESEARCH","world_novelty_boundary":"Assurance records, claim-evidence-argument structures, pre-award supplier review, independent testing, contractual acceptance conditions, monitoring, and change-triggered reassessment are established or directly documented practices; bounded-search novelty remains only for making computability theory itself—input class, quantifiers, computation model, decidability/recognizability/unknown status, totality evidence, and reduction validity—a mandatory procurement subcase. The search found no direct implementation of that narrow combination, but it also found no evidence that unsupported universal exact terminating guarantees are a measured recurring procurement failure, so world novelty and practical need remain unproven."}