{"schema_version":1,"assessment_id":"eoa_inverse_innovation_exp03_opportunity320_20260801","source_experiment_id":"eoa_inverse_innovation_exp03_full320_20260801","cell_id":"computability_boundary_mapping__rhetoric","archetype_slug":"computability_boundary_mapping","domain_slug":"rhetoric","title":"Scoped Reachability Guarantees for Executable Persuasive Dialogue","opportunity_summary":"Conditionally replace unrestricted Boolean safety claims with an enforced finite campaign fragment that admits exact checking, while routing other campaigns to bounded witness search or explicit UNKNOWN. The opportunity depends on verifying that the platform actually admits unbounded computation and that formal audience-model reachability is not represented as real-world persuasion safety.","adopter_authorizer":"The rhetoric platform's safety owner can authorize offline validation; governance leaders and affected-party representatives must authorize deployment claims.","scores":{"meaningful_impact":{"score":4,"rationale":"If the described unrestricted analyzer is being pursued, preventing timeouts from being treated as SAFE could reduce false clearances, false blocks, and wasted engineering effort. The frequency and reach of such failures are not supplied."},"stakeholder_pull":{"score":3,"rationale":"Platform designers, moderation analysts, campaign authors, and audiences have identifiable interests in trustworthy verdicts, but the packet supplies no evidence of demand, current incidents, procurement intent, or willingness to accept restrictions and UNKNOWN results."},"incremental_advantage":{"score":4,"rationale":"Enforced fragment membership, total checking within that fragment, and explicit UNKNOWN routing provide a clear auditable advantage over timeout-as-safe Boolean analysis and over a confidence-scored heuristic lacking a guarantee boundary. Practical utility could fall if UNKNOWN rates are high."},"distinctiveness_plausibility":{"score":3,"rationale":"The composition of a checked reduction, enforceable decidable fragment, router, scoped labels, and rechecking is coherent, but prior art is explicitly unsearched, so distinctiveness cannot be established closed-book."},"technical_implementability":{"score":3,"rationale":"A synthetic finite-state checker and labeled router appear technically buildable, but the unrestricted-language expressiveness premise, reduction, abstraction fidelity, fragment enforcement, and production integration remain unverified obligations."},"adoption_authority_feasibility":{"score":4,"rationale":"The proposal names a platform safety owner for offline work and governance plus affected-party representatives for deployment claims, with explicit exclusions and rollback conditions. Multi-party approval and enforceable language restrictions may still complicate adoption."},"evidence_readiness":{"score":3,"rationale":"The packet supplies strong falsifiers, negative tests, and a bounded synthetic pilot design, but no checked reduction, language audit, prototype results, operational data, or prior-art review."},"safety_net_benefit":{"score":5,"rationale":"Explicit UNKNOWN outcomes, prohibition on treating UNKNOWN as SAFE, synthetic-only first testing, guarantee labels, manual quarantine, and concrete halt conditions directly limit false assurance while evidence is developed."},"scalability":{"score":3,"rationale":"The fragment-and-router pattern could be reused across campaigns, but checker complexity, potentially high UNKNOWN rates, author displacement, model-specific abstractions, and required rechecking after language changes make scale uncertain."}},"score_confidence":"MODERATE","costs":{"first_evidence":{"band_2026_usd":"50K_TO_250K","scope":"A bounded offline language audit, formalization of reachability and verdict semantics, independently reviewed proof obligations, finite-state checker prototype, synthetic corpus construction, seeded harmful traces, and comparison with timeout-as-safe and heuristic baselines.","confidence":"LOW","assumptions":["A documented campaign language and audience-state semantics are available to the research team.","No real audiences or production campaigns are used.","The study covers one language and one formal audience model.","Specialist formal-methods and rhetoric-safety labor is included."]},"initial_deployment_startup":{"band_2026_usd":"250K_TO_1M","scope":"Production-grade fragment validator, exact checker, out-of-scope router, scoped result labels, audit logging, adversarial testing, governance review, analyst workflow integration, and rollback controls for one platform.","confidence":"LOW","assumptions":["The offline study supports proceeding.","Existing platform interfaces can expose campaign syntax, histories, and model versions.","Deployment does not require rebuilding the entire campaign-authoring system.","Legal, compliance, accessibility, and affected-party review are included."]},"operational_launch":{"band_2026_usd":"1M_TO_5M","scope":"Controlled platform launch with migration of admitted campaigns, author and analyst training, monitoring of UNKNOWN and false-alarm burdens, independent assurance, incident response, and enforcement against fragment bypass.","confidence":"LOW","assumptions":["Launch spans a materially used platform rather than a laboratory environment.","Manual quarantine capacity is maintained for UNKNOWN cases.","Formal-model guarantees are clearly separated from claims about real audience response.","No live persuasion experiment is conducted as part of validation."]},"annual_recurring":{"band_2026_usd":"250K_TO_1M","scope":"Ongoing checker maintenance, model and language change review, periodic proof and implementation audits, synthetic regression testing, governance reporting, label monitoring, and manual escalation support.","confidence":"LOW","assumptions":["One principal platform and a limited number of language or model revisions are supported.","Major semantic redesigns are treated as new startup work.","UNKNOWN cases do not require a large permanent moderation workforce.","Independent review is repeated after material expressiveness changes."]}},"research_burden":"HIGH","earliest_credible_horizon":"3_TO_12_MONTHS","pipeline_gates":{"recognizable_externally_supportable_problem":{"status":"UNCERTAIN","reason":"The packet describes a coherent failure—an exact terminating analyzer over unrestricted executable campaigns with timeout treated as SAFE—but also states that the required unbounded expressiveness premise needs local verification and supplies no evidence that the deployed class is not already finite."},"identifiable_adopter_or_authorizer":{"status":"YES","reason":"The platform safety owner is identified for offline authorization, with governance and affected-party representatives assigned authority over deployment claims."},"distinct_testable_incremental_claim":{"status":"YES","reason":"The proposal claims that enforced finite-fragment checking plus explicit UNKNOWN routing will avoid incorrect unrestricted Boolean guarantees; termination, verdict correctness, membership enforcement, and routing labels are directly testable against the stated baselines."},"bounded_next_evidence_step":{"status":"YES","reason":"The sealed candidate authorizes an offline synthetic pilot using small exhaustive instances and seeded harmful traces, with explicit comparison conditions and failure criteria and no exposure of real audiences."},"no_unresolved_safety_or_authority_stop":{"status":"YES","reason":"The first step is synthetic and offline, excludes live persuasion and automatic clearance, prohibits interpreting UNKNOWN as SAFE, and defines halt and rollback triggers. Deployment remains separately gated."},"implementation_cost_scope_and_range":{"status":"YES","reason":"A one-language offline study and a one-platform deployment can be bounded into broad resource bands, although costs remain low-confidence until language semantics, integration surfaces, and expected UNKNOWN volume are inspected."}},"blocking_evidence":["Whether all admitted campaigns and histories are already effectively finite under enforced bounds.","Whether the unrestricted campaign language can simulate arbitrary computation under the platform's actual semantics.","An independently checked, total, computable, answer-preserving reduction if an undecidability claim is pursued.","Termination and verdict correctness of the finite-fragment checker, including seeded harmful traces.","Enforceability of fragment membership and correct routing of every out-of-scope input to non-exact treatment.","Abstraction fidelity between campaign behavior and the stipulated audience-state model.","Operational UNKNOWN rates and their burden on manual quarantine and legitimate advocacy.","Prior-art evidence concerning dialogue-system verification, executable campaign languages, and abstaining moderation architectures."],"next_evidence_step":"Run a time-bounded offline study on one documented campaign language and one synthetic finite-state audience model. First audit whether enforced bounds already falsify the unrestricted-computability premise. Then prototype the fragment checker and router, exhaust all instances up to declared bounds, seed known harmful traces, and compare results with timeout-as-safe Boolean analysis and the confidence-scored heuristic rival. Stop if any harmful bounded trace is cleared, any well-formed in-fragment case fails to terminate or receives an incorrect verdict, or any out-of-scope case is labeled exact rather than UNKNOWN.","research_questions":["Are admitted scripts, audience-state updates, and interaction histories genuinely unbounded, or are effective finite limits already enforced?","If the language is computationally expressive, can independent reviewers validate the proposed reachability reduction and its preservation mapping?","Which rhetorically necessary interactions are excluded by the decidable fragment?","Can fragment membership be enforced across authoring, compilation, model updates, and execution without bypasses?","What proportion of representative campaigns routes to UNKNOWN, and can moderation operations safely handle that volume?","Do result labels prevent analysts and authors from interpreting formal-model reachability as evidence of real audience safety?","How does the checker compare with the baseline and nearest rival on seeded harmful traces, benign cases, termination, and abstention behavior?","What relevant prior art limits claims of distinctiveness or changes the appropriate implementation design?"],"recommendation":"PARTNERED_RESEARCH","uncertainty_constraints":["Problem prevalence, market size, stakeholder demand, realized harm, and willingness to adopt are unmeasured.","The central undecidability conclusion is conditional on locally unverified computational expressiveness.","The deployed campaign class may already be finite, making complexity rather than computability the relevant issue.","Formal audience-model reachability cannot establish actual human persuasion outcomes.","Prior art and world-level distinctiveness are unknown.","Cost bands are resource-equivalent planning ranges, not observed vendor or staffing prices.","Operational usefulness depends on fragment coverage, UNKNOWN rates, abstraction fidelity, and governance behavior."],"closed_book_prior_art_boundary":"No claim is made about novelty, prevalence, market position, existing implementations, or realized impact. The packet explicitly marks prior art as unsearched; all distinctiveness judgments are therefore provisional and limited to the internal structure of the sealed candidate."}