{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp03_full320_20260801","cell_id":"computability_boundary_mapping__veterinary_medicine","trajectory_id":"R","attempt_index":0,"archetype_slug":"computability_boundary_mapping","domain_slug":"veterinary_medicine","decision":"CANDIDATE","problem_id":"universal_veterinary_welfare_controller_safety_decision","causal_lever_id":"restrict_and_label_welfare_safety_analysis_guarantees","proposal":{"problem":"A veterinary organization seeks an exact, always-terminating predeployment analyzer that decides, for every closed-loop animal-monitoring or treatment controller expressible in an unrestricted executable rule language, whether any future execution can violate a stated animal-welfare constraint. Timeouts and successful simulations are liable to be reported as Boolean safety verdicts even though the class, environment, and guarantee have not been bounded.","actors_substrate":["animals subject to automated monitoring or treatment","veterinarians and animal-behavior specialists","controller and sensor vendors","animal-care operators","software assurance reviewers"],"observable_state":"Safety reports collapse verified-safe, violation-found, timeout, model-mismatch, and out-of-scope cases into pass/fail while unrestricted controllers and unbounded interaction histories remain admitted.","consequence":"An impossible universal analyzer may absorb development effort, while false safety clearance or false rejection can expose animals to harm or withhold useful care.","affected_objective":"Obtain trustworthy predeployment welfare assurance without overstating what analysis of an executable care controller can decide.","structural_mapping":[{"archetype_element":"unrestricted universal decision request","domain_realization":"Decide welfare reachability for every executable controller over every future animal-environment interaction.","claim_kind":"HYPOTHESIS"},{"archetype_element":"explicit input and computation model","domain_realization":"Specify controller syntax, sensor values, actuator actions, physiology abstraction, environment transitions, and available expert oracles.","claim_kind":"INFERENCE"},{"archetype_element":"computability boundary","domain_realization":"Separate finite-state controllers and bounded traces from unrestricted controllers with unbounded state or execution.","claim_kind":"INFERENCE"},{"archetype_element":"constructive and impossibility evidence","domain_realization":"Require a terminating decider for restricted classes or a valid source-to-target reduction before declaring the unrestricted class undecidable.","claim_kind":"INFERENCE"},{"archetype_element":"honest weaker fallback","domain_realization":"Route eligible cases to exact checking, sound abstraction, bounded search, or explicit UNKNOWN and veterinary review.","claim_kind":"INFERENCE"},{"archetype_element":"versioned reclassification","domain_realization":"Reassess guarantees when controller syntax, welfare predicates, sensors, environment assumptions, or external expertise change.","claim_kind":"INFERENCE"}],"component_map":[{"component":"Problem-Class Specification","status":"adapted","domain_realization":"Controllers and welfare-reachability questions covered."},{"component":"Instance Representation Contract","status":"adapted","domain_realization":"Machine-checkable controller, state, transition, and welfare-predicate encoding."},{"component":"Computation Model Contract","status":"adapted","domain_realization":"Declared machine, interaction, randomness, sensors, and expert calls."},{"component":"Solvability Guarantee Profile","status":"direct","domain_realization":"Exact, sound-one-sided, bounded, or unresolved verdict."},{"component":"Quantifier and Scope Map","status":"direct","domain_realization":"Every controller, execution, horizon, and admissible environment made explicit."},{"component":"Computability Status Lattice","status":"direct","domain_realization":"Decidable, recognizable, relative, undecidable, or unresolved."},{"component":"Constructive Procedure Witness","status":"adapted","domain_realization":"Algorithm plus correctness and termination proof for an admitted fragment."},{"component":"Reduction Preservation Contract","status":"adapted","domain_realization":"Computable source-to-controller translation preserving the welfare answer."},{"component":"Computability Impossibility Certificate","status":"adapted","domain_realization":"Checked reduction or direct proof for the exact unrestricted class."},{"component":"Assumption Register","status":"direct","domain_realization":"Encoding, dynamics, sensor, oracle, and welfare-semantics assumptions."},{"component":"Decidable Subclass Map","status":"adapted","domain_realization":"Finite-state, syntactically restricted, promised, and bounded-horizon classes."},{"component":"One-Sided Recognition Contract","status":"adapted","domain_realization":"Confirmed violation requires a checkable trace; non-discovery is not safety."},{"component":"Unknown and Nontermination Policy","status":"direct","domain_realization":"UNKNOWN, timeout, out-of-scope, and failure remain distinct."},{"component":"Fallback Solution Contract","status":"adapted","domain_realization":"Guarantee-labelled exact, abstract, bounded, or reviewed mode."},{"component":"Computability Guarantee Record","status":"direct","domain_realization":"Versioned shipped claim linked to evidence."},{"component":"Recheck Trigger","status":"adapted","domain_realization":"Language, model, predicate, sensor, or oracle change."},{"component":"Termination Condition","status":"direct","domain_realization":"Finite-state completion or declared resource bound."},{"component":"Scope Boundary","status":"direct","domain_realization":"Enforceable fragment, promise, and horizon limits."},{"component":"Decision Record","status":"direct","domain_realization":"Chosen boundary, rationale, owner, and expiry conditions."},{"component":"Uncertainty Residue","status":"direct","domain_realization":"Open proof obligations and model-to-animal gaps."},{"component":"Independent Proof Review","status":"adapted","domain_realization":"Independent checking of proofs, reductions, and formalized claims."},{"component":"Complexity Follow-On Gate","status":"direct","domain_realization":"Decidable classes proceed to runtime and state-space assessment."}],"mechanism_dispositions":[{"slug":"abstract_interpretation_or_model_checking","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Over-approximate reachable welfare states; only abstraction-safe clearance is conclusive.","counterfactual_removal":"No sound scalable fallback for large finite models."},{"slug":"bounded_domain_exhaustive_search","disposition":"selected_supporting","contribution_type":"TEST_DESIGN","adaptation_or_rejection":"Enumerate all executions within explicit state and horizon bounds.","counterfactual_removal":"Bounded pilot loses complete within-bound evidence."},{"slug":"computability_boundary_decision_record","disposition":"selected_supporting","contribution_type":"OPERATIONAL","adaptation_or_rejection":"Version the class, evidence, shipped claim, and triggers.","counterfactual_removal":"Guarantees can drift silently."},{"slug":"computational_complexity_analysis","disposition":"selected_supporting","contribution_type":"OPERATIONAL","adaptation_or_rejection":"Assess feasibility only after decidability classification.","counterfactual_removal":"Computable fragments may still be operationally unusable."},{"slug":"constructive_algorithm_and_correctness_proof","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Certify total exact analyzers for restricted classes.","counterfactual_removal":"Decidable placement lacks a witness."},{"slug":"diagonalization_impossibility_proof","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"A domain-specific reduction is clearer; direct self-reference is unnecessary.","counterfactual_removal":"No change if reduction evidence exists."},{"slug":"enumeration_and_dovetailing","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Violation-trace search is better governed by bounded explicit-UNKNOWN operation.","counterfactual_removal":"No change to the bounded interface."},{"slug":"fallback_mode_router","disposition":"selected_load_bearing","contribution_type":"OPERATIONAL","adaptation_or_rejection":"Dispatch by enforceable class and attach the applicable guarantee.","counterfactual_removal":"Weak outputs may be laundered as exact safety verdicts."},{"slug":"halting_problem_reduction","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Attempt a computable embedding only if unrestricted controllers can simulate arbitrary computation.","counterfactual_removal":"No justified unrestricted undecidability conclusion."},{"slug":"language_fragment_restriction","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Admit a mechanically enforceable finite-state controller fragment.","counterfactual_removal":"Exact terminating service lacks a safe admissible class."},{"slug":"many_one_reduction_proof","disposition":"selected_supporting","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Supplies totality and answer-preservation obligations for the halting reduction.","counterfactual_removal":"Reduction may remain rhetorical or directionally invalid."},{"slug":"promise_problem_restriction","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Unenforced physiological promises permit authoritative wrong answers; prefer syntactic admission.","counterfactual_removal":"No loss under fragment enforcement."},{"slug":"proof_by_counterexample","disposition":"selected_supporting","contribution_type":"TEST_DESIGN","adaptation_or_rejection":"Refute an existing universal analyzer claim with one admitted failing controller.","counterfactual_removal":"Overbroad claims become harder to retire cheaply."},{"slug":"proof_checking","disposition":"selected_load_bearing","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"Independently check the formal theorem and premises.","counterfactual_removal":"A proof gap could authorize unsafe clearance."},{"slug":"reduction_direction_checklist","disposition":"selected_supporting","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"Gate source-to-target direction and declared assumptions.","counterfactual_removal":"Hardness may be transferred backward."},{"slug":"semi_decision_with_explicit_unknown","disposition":"selected_load_bearing","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"Return witnessed VIOLATION or bounded UNKNOWN, never timeout-as-safe.","counterfactual_removal":"Non-discovery can become false safety."},{"slug":"theorem_prover_guided_search","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Optional implementation aid; proof search is not needed for the first classification.","counterfactual_removal":"Core evidence obligations remain."},{"slug":"turing_reduction_analysis","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Relative oracle degrees do not answer the initial shipped-guarantee question.","counterfactual_removal":"No material change."}],"causal_chain":["Formalize the controller class, welfare property, quantifiers, and computation model.","Test whether the unrestricted language supports a valid undecidability reduction; otherwise mark unresolved.","Enforce a decidable fragment and explicit finite bounds where exact termination is required.","Route each admitted query to exact, sound-abstract, bounded, or UNKNOWN output.","Independent review and versioned records prevent stronger claims than the evidence supports."],"baseline":"Run simulations and tests until timeout, then issue an informal pass/fail judgment with veterinarian escalation handled ad hoc.","nearest_rival":"Treat the task solely as state-space or runtime optimization while retaining the universal exact safety claim.","authority_safety":{"affected_parties":["animals receiving automated care","animal owners or custodians","veterinary staff","care-facility operators"],"decision_authority":"The responsible veterinarian retains clinical authority; an animal-welfare or ethics body approves safety claims, and software assurance approves formal evidence.","authorized_first_step":"Offline shadow audit of up to 30 de-identified controller configurations: formalize one welfare predicate, classify their language and guarantees, and compare labels with existing reports without changing animal care.","excluded_actions":["autonomous treatment changes","declaring a controller clinically safe from model evidence alone","interpreting UNKNOWN or timeout as safe","expanding data collection or animal experimentation"],"halt_rollback":"Halt the pilot and withdraw its labels if an observed transition is absent from the model, an unsafe case receives a safe label, UNKNOWN is coerced to pass/fail, or review workload exceeds the approved bound; preserve existing clinical workflow."}},"negative_tests":{"strongest_counterevidence":"If every deployed controller and environment model is finite, effectively enumerable, and bounded by enforceable limits, the task is decidable in principle and the real problem is complexity rather than computability.","analogy_break":"Animal physiology and behavior are not executable programs. A theorem about an encoded controller establishes only model-relative reachability; it cannot by itself establish real-world clinical safety or complete environment fidelity.","failure_condition":"Fragment restrictions exclude routine care protocols, abstractions generate unusable false alarms, or vendors can bypass admission checks while outputs still appear authoritative.","problem_falsifier":"An audit finds no class-wide exact-and-terminating claim: all requested judgments are explicitly bounded instances with distinct UNKNOWN and clinical-review states.","intervention_falsifier":"After the bounded audit, unsupported Boolean safety claims, timeout-as-safe decisions, or investment in an unrestricted total analyzer persist at the same rate, or independent reviewers cannot reproduce classifications.","risks":["False confidence from an unsound or clinically unfaithful abstraction","Loss of useful protocols through an overly narrow fragment","Automation bias toward formally labelled outputs","Delayed care from excessive UNKNOWN results","A correct computability result being mistaken for evidence of practical feasibility"]},"null_rationale":null,"classification":{"candidate_kind":"DOMAIN_TRANSFER","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 candidate depends on the controller language genuinely being unrestricted; absent that fact, undecidability remains a testable hypothesis rather than a domain claim."}