{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp03_full320_20260801","cell_id":"computability_boundary_mapping__speech_language_pathology","trajectory_id":"R","attempt_index":0,"archetype_slug":"computability_boundary_mapping","domain_slug":"speech_language_pathology","decision":"CANDIDATE","problem_id":"unverifiable_unrestricted_aac_dialogue_policies","causal_lever_id":"enforceable_aac_policy_language_boundary","proposal":{"problem":"A speech-language pathology service deploys clinician-authored, adaptive AAC dialogue policies while claiming that an automated verifier can always determine whether every policy will avoid prohibited utterances and eventually offer a usable communication path for every possible interaction history. If the policy language permits unbounded state or computation, that unrestricted total-exact verification claim may be impossible; timeouts are currently liable to be interpreted as safety or failure.","actors_substrate":["AAC users and communication partners","speech-language pathologists","caregivers and clinical governance staff","AAC policy authors and software maintainers","programmable AAC policy language, verifier, and runtime"],"observable_state":"Verification jobs time out; results lack distinct false, unknown, and out-of-scope states; documentation makes unrestricted universal claims; policy-language extensions are not tied to guarantee reclassification.","consequence":"An unsafe or nonterminating policy may be cleared, a safe policy may be withheld, or a user may lose timely access to intended communication.","affected_objective":"Preserve communicative agency and safety while publishing only verification guarantees justified for the enforceable policy class.","structural_mapping":[{"archetype_element":"Unrestricted problem class","domain_realization":"All accepted AAC policies across all interaction histories and user inputs.","claim_kind":"INFERENCE"},{"archetype_element":"Total exact decider demand","domain_realization":"The verifier must halt and correctly answer both safety and eventual-path questions for every accepted policy.","claim_kind":"INFERENCE"},{"archetype_element":"Computability boundary","domain_realization":"Expressiveness of the AAC policy language determines whether universal semantic verification remains decidable.","claim_kind":"HYPOTHESIS"},{"archetype_element":"Restricted decidable region","domain_realization":"A mechanically enforced finite-state or otherwise proved-decidable policy fragment.","claim_kind":"HYPOTHESIS"},{"archetype_element":"Governed fallback","domain_realization":"Exact verification inside the fragment; sound abstraction, bounded analysis, or UNKNOWN outside stronger guarantees.","claim_kind":"INFERENCE"}],"component_map":[{"component":"Problem-Class Specification","status":"adapted","domain_realization":"Define policy properties, histories, and universal coverage."},{"component":"Instance Representation Contract","status":"adapted","domain_realization":"Versioned policy syntax, state, inputs, and runtime semantics."},{"component":"Computation Model Contract","status":"adapted","domain_realization":"State verifier capabilities, bounds, and any human or service oracle."},{"component":"Solvability Guarantee Profile","status":"adapted","domain_realization":"Separate exact, sound-incomplete, bounded, and unresolved guarantees."},{"component":"Quantifier and Scope Map","status":"adapted","domain_realization":"Distinguish one policy, bounded histories, and all policies/histories."},{"component":"Computability Status Lattice","status":"adapted","domain_realization":"Classify fragments as decidable, recognizable, relative, or unresolved."},{"component":"Constructive Procedure Witness","status":"adapted","domain_realization":"Provide a terminating correct verifier for each exact fragment."},{"component":"Reduction Preservation Contract","status":"adapted","domain_realization":"Require a total semantics-preserving embedding before claiming undecidability."},{"component":"Computability Impossibility Certificate","status":"adapted","domain_realization":"Condition any unrestricted impossibility verdict on a checked reduction."},{"component":"Assumption Register","status":"direct","domain_realization":"Record language, runtime, environment, and property assumptions."},{"component":"Decidable Subclass Map","status":"adapted","domain_realization":"List enforceable policy fragments and their supported properties."},{"component":"One-Sided Recognition Contract","status":"adapted","domain_realization":"Identify witness-bearing findings that can be confirmed without complete negatives."},{"component":"Unknown and Nontermination Policy","status":"adapted","domain_realization":"Keep UNKNOWN, timeout, false, out-of-scope, and tool failure distinct."},{"component":"Fallback Solution Contract","status":"adapted","domain_realization":"Label every result with the method and guarantee used."},{"component":"Computability Guarantee Record","status":"direct","domain_realization":"Version-link each shipped claim to evidence and policy-language version."},{"component":"Recheck Trigger","status":"adapted","domain_realization":"Reclassify after syntax, semantics, property, or oracle changes."},{"component":"Termination Condition","status":"adapted","domain_realization":"Declare exhaustive bounds and analysis budgets."},{"component":"Scope Boundary","status":"adapted","domain_realization":"Mechanically reject or quarantine policies outside exact fragments."},{"component":"Decision Record","status":"direct","domain_realization":"Record the chosen verification boundary and rationale."},{"component":"Uncertainty Residue","status":"adapted","domain_realization":"Retain unproved obligations and abstraction alarms."},{"component":"Independent Proof Review","status":"direct","domain_realization":"Separate review of reductions and verifier proofs."},{"component":"Complexity Follow-On Gate","status":"adapted","domain_realization":"Assess feasibility only after establishing decidability."}],"mechanism_dispositions":[{"slug":"abstract_interpretation_or_model_checking","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Sound finite abstraction for selected safety properties; alarms remain unresolved.","counterfactual_removal":"Removes the principal sound fallback for policies outside exact fragments."},{"slug":"bounded_domain_exhaustive_search","disposition":"selected_supporting","contribution_type":"TEST_DESIGN","adaptation_or_rejection":"Exhaust synthetic policies and histories only within declared bounds.","counterfactual_removal":"Weakens bounded validation but not the boundary classification."},{"slug":"computability_boundary_decision_record","disposition":"selected_load_bearing","contribution_type":"OPERATIONAL","adaptation_or_rejection":"Version the boundary, guarantee, assumptions, and triggers.","counterfactual_removal":"Guarantees can drift silently as the language changes."},{"slug":"computational_complexity_analysis","disposition":"selected_supporting","contribution_type":"OPERATIONAL","adaptation_or_rejection":"Estimate cost after a fragment is shown decidable.","counterfactual_removal":"Leaves practical deployment feasibility unresolved."},{"slug":"constructive_algorithm_and_correctness_proof","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Required for any fragment advertised as exactly decidable.","counterfactual_removal":"The exact-fragment claim lacks a total correct witness."},{"slug":"diagonalization_impossibility_proof","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Redundant unless a direct self-reference construction fits the policy semantics.","counterfactual_removal":"No change; reduction evidence is sufficient."},{"slug":"enumeration_and_dovetailing","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Unbounded nontermination is unsuitable for a clinical release gate.","counterfactual_removal":"No change; bounded explicit-UNKNOWN protocol is used."},{"slug":"fallback_mode_router","disposition":"selected_load_bearing","contribution_type":"OPERATIONAL","adaptation_or_rejection":"Route by fragment and label exact, abstract, bounded, or escalated results.","counterfactual_removal":"Weaker results may be mistaken for exact clearance."},{"slug":"halting_problem_reduction","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Attempt only if the unrestricted policy language can faithfully encode source instances.","counterfactual_removal":"No principled basis remains for an undecidability verdict."},{"slug":"language_fragment_restriction","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Create a syntactically enforceable exact-verification fragment.","counterfactual_removal":"The service retains an unrestricted guarantee it may not support."},{"slug":"many_one_reduction_proof","disposition":"selected_supporting","contribution_type":"TEST_DESIGN","adaptation_or_rejection":"Formalize totality, computability, and answer preservation of the impossibility mapping.","counterfactual_removal":"The reduction certificate is less auditable."},{"slug":"promise_problem_restriction","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"A mere caller promise is unsafe unless mechanically enforced; use syntax restriction.","counterfactual_removal":"No change."},{"slug":"proof_by_counterexample","disposition":"selected_supporting","contribution_type":"TEST_DESIGN","adaptation_or_rejection":"Refute overbroad verifier claims with an in-scope failing policy.","counterfactual_removal":"Universal overclaims become harder to falsify cheaply."},{"slug":"proof_checking","disposition":"selected_load_bearing","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"Independent review of fragment proofs and reductions.","counterfactual_removal":"A proof gap could authorize a false clinical guarantee."},{"slug":"reduction_direction_checklist","disposition":"selected_supporting","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"Gate any source-to-target hardness claim.","counterfactual_removal":"Direction errors become easier to institutionalize."},{"slug":"semi_decision_with_explicit_unknown","disposition":"selected_load_bearing","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"Bound one-sided searches and return UNKNOWN, never fabricated NO.","counterfactual_removal":"Timeouts can again masquerade as verdicts."},{"slug":"theorem_prover_guided_search","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Optional tooling, not necessary for the first boundary test.","counterfactual_removal":"Manual independent proofs remain possible."},{"slug":"turing_reduction_analysis","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Relative computability degrees do not answer the immediate release decision.","counterfactual_removal":"No change."}],"causal_chain":["Specify accepted AAC policies, semantics, properties, and universal quantifiers.","Test whether unrestricted policies admit a checked constructive decider or valid impossibility reduction.","Mechanically define exact decidable fragments and retain unresolved classifications where evidence fails.","Route each policy to exact, sound-abstract, bounded-UNKNOWN, or human-review handling.","Expose guarantee labels and prevent UNKNOWN or timeout from authorizing deployment."],"baseline":"Run the current verifier on each policy; retry timeouts or treat them informally as failure/safety, with expert review by exception.","nearest_rival":"Keep the unrestricted language and use testing plus a larger timeout, without changing the universal guarantee.","authority_safety":{"affected_parties":["AAC users","communication partners","speech-language pathologists","caregivers","policy authors"],"decision_authority":"Clinical AAC governance lead jointly with the affected user's authorized care team and software safety owner; the user retains control over personal communication choices.","authorized_first_step":"Retrospective sandbox pilot on synthetic policies and de-identified policy structures: formalize one safety property, one bounded history model, and one candidate restricted fragment; compare result labels without controlling a live device.","excluded_actions":["No live-policy blocking based solely on UNKNOWN","No automatic replacement of a user's vocabulary or intent","No claim of undecidability without a checked semantics-preserving proof","No silent interpretation of timeout as safe or false"],"halt_rollback":"Stop if the model omits clinically material runtime behavior, an in-bound exact verdict is wrong, UNKNOWN is consumed as clearance, or users lose access; revert to existing human release review and withdraw the affected guarantee."}},"negative_tests":{"strongest_counterevidence":"The deployed policy language and interaction horizon may already be finite, mechanically bounded, and exhaustively enumerable; then the problem is decidable and primarily one of complexity or implementation.","analogy_break":"Program properties map to AAC policies only if the policy syntax and runtime semantics can express the computations used in the proof; natural conversational openness alone does not establish undecidability.","failure_condition":"The restricted fragment excludes common communication needs, membership is not enforced, abstractions omit real behavior, or downstream systems collapse UNKNOWN into approval.","problem_falsifier":"No unrestricted programmable policy class or universal total-exact verification claim exists; accepted policies are finitely bounded and the observed issue is only runtime cost.","intervention_falsifier":"On a preregistered bounded corpus, the boundary workflow misclassifies an in-fragment policy, fails to terminate within its stated exact bound, or emits labels that reviewers cannot reliably distinguish.","risks":["False reassurance from an unsound abstraction","Loss of AAC expressiveness or user agency","Excessive UNKNOWN results delaying access","A formally correct model that misrepresents clinical use","Governance records preserving a mistaken boundary"]},"null_rationale":null,"classification":{"candidate_kind":"MECHANISM_COMPOSITION","prior_art_status":"UNSEARCHED","evidence_maturity":"HYPOTHESIS"},"revision_change_log":{"revision_kind":"ORIGINAL","prior_problem_id":null,"prior_causal_lever_id":null,"problem_changed":false,"causal_lever_changed":false,"conceptual_changes":[],"operational_changes":[],"repairs_addressed":[]},"confidence":0.78,"generator_notes":"Closed-book structural transfer. The computability claim is conditional on the AAC policy language and runtime supporting a valid encoding; the packet provides no evidence that a deployed system actually has that expressiveness."}