{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp03_full320_20260801","cell_id":"computability_boundary_mapping__pharmacology_toxicology","trajectory_id":"R","attempt_index":0,"archetype_slug":"computability_boundary_mapping","domain_slug":"pharmacology_toxicology","decision":"CANDIDATE","problem_id":"universal_executable_toxicity_model_safety_decision","causal_lever_id":"model_relative_toxicity_reachability_boundary","proposal":{"problem":"A computational-toxicology project claims an exact, always-terminating analyzer that decides whether any arbitrarily expressive executable PK/PD or toxicodynamic model can ever produce a specified toxic endpoint over unbounded execution. Failed searches, timeouts, and results on restricted models are liable to be reported as universal safety verdicts. [HYPOTHESIS]","actors_substrate":["computational toxicologists and pharmacometricians","model and analyzer developers","safety reviewers and decision-makers","patients, trial participants, exposed populations, and animals affected by downstream decisions","executable biological models, endpoint specifications, exposure inputs, and analysis records"],"observable_state":"The interface emits SAFE/TOXIC for every submitted model, while accepted model syntax, horizon, nondeterminism, external information, termination guarantee, and meaning of timeout are undocumented or inconsistent. [HYPOTHESIS]","consequence":"An impossible unrestricted guarantee can consume development effort or convert UNKNOWN into false reassurance; conversely, an unjustified impossibility claim can suppress useful analysis of bounded or decidable model classes. [INFERENCE]","affected_objective":"Produce auditable toxicity-model verdicts whose scope, termination, soundness, completeness, and practical feasibility match the models actually analyzed.","structural_mapping":[{"archetype_element":"open-ended problem class","domain_realization":"Arbitrary executable PK/PD or toxicodynamic models queried for eventual toxic-endpoint reachability","claim_kind":"HYPOTHESIS"},{"archetype_element":"total exact decision requirement","domain_realization":"Return the correct SAFE/TOXIC answer and terminate for every accepted model","claim_kind":"HYPOTHESIS"},{"archetype_element":"computability-model boundary","domain_realization":"Reachability is classified relative to model language, input encoding, horizon, nondeterminism, and declared external capabilities","claim_kind":"INFERENCE"},{"archetype_element":"impossibility evidence","domain_realization":"A checked source-to-target halting reduction embeds program termination as toxic-endpoint reachability","claim_kind":"HYPOTHESIS"},{"archetype_element":"decidable region","domain_realization":"Enforceable finite-state, bounded-horizon, or otherwise total restricted model fragments","claim_kind":"INFERENCE"},{"archetype_element":"honest fallback","domain_realization":"Sound abstraction, bounded checking, or witness search returns guarantee-labelled TOXIC, SAFE-UNDER-CONTRACT, or UNKNOWN","claim_kind":"INFERENCE"}],"component_map":[{"component":"Problem-Class Specification","status":"adapted","domain_realization":"Define the accepted toxicity-model class and endpoint-reachability question."},{"component":"Instance Representation Contract","status":"adapted","domain_realization":"Specify model syntax, parameters, exposure inputs, endpoint predicate, and horizon encoding."},{"component":"Computation Model Contract","status":"adapted","domain_realization":"Declare execution semantics, precision, nondeterminism, resources, and external services."},{"component":"Solvability Guarantee Profile","status":"direct","domain_realization":"State termination, soundness, completeness, and scope per analysis mode."},{"component":"Quantifier and Scope Map","status":"direct","domain_realization":"Separate every-model claims from one-model, bounded-horizon, and restricted-fragment claims."},{"component":"Computability Status Lattice","status":"direct","domain_realization":"Classify modes as decidable, recognizable, partial, relative, or unresolved."},{"component":"Constructive Procedure Witness","status":"adapted","domain_realization":"For each decidable fragment, provide an analyzer with correctness and termination arguments."},{"component":"Reduction Preservation Contract","status":"adapted","domain_realization":"Show the halting-instance encoding is total, computable, in scope, and answer-preserving."},{"component":"Computability Impossibility Certificate","status":"adapted","domain_realization":"Retain the checked reduction proving only the matched unrestricted claim."},{"component":"Assumption Register","status":"direct","domain_realization":"Record semantics, encoding, endpoint, horizon, and capability assumptions."},{"component":"Decidable Subclass Map","status":"adapted","domain_realization":"Map enforceable finite-state, bounded, and restricted-language model regions."},{"component":"One-Sided Recognition Contract","status":"adapted","domain_realization":"Define which toxic witnesses can be confirmed without claiming universal safe negatives."},{"component":"Unknown and Nontermination Policy","status":"direct","domain_realization":"Keep timeout, UNKNOWN, out-of-scope, SAFE, and TOXIC distinct."},{"component":"Fallback Solution Contract","status":"adapted","domain_realization":"Label abstraction, bounded search, and escalation outputs with their exact guarantees."},{"component":"Computability Guarantee Record","status":"direct","domain_realization":"Version-link each shipped claim to model class, evidence, and analyzer."},{"component":"Recheck Trigger","status":"adapted","domain_realization":"Reclassify after language, semantics, endpoint, horizon, or external-capability changes."},{"component":"Termination Condition","status":"direct","domain_realization":"Set explicit finite bounds for operational fallback runs."},{"component":"Scope Boundary","status":"direct","domain_realization":"Mechanically reject or quarantine inputs outside guaranteed fragments."},{"component":"Decision Record","status":"direct","domain_realization":"Record the selected boundary, rationale, fallback, and supersession history."},{"component":"Uncertainty Residue","status":"adapted","domain_realization":"List unproved obligations, abstraction imprecision, and unresolved model classes."},{"component":"Independent Proof Review","status":"direct","domain_realization":"Have an independent reviewer check theorem match and reduction details."},{"component":"Complexity Follow-On Gate","status":"direct","domain_realization":"After decidability, assess state-space growth and operational feasibility."}],"mechanism_dispositions":[{"slug":"abstract_interpretation_or_model_checking","disposition":"selected_supporting","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Use sound finite abstractions for guarantee-labelled restricted safety results.","counterfactual_removal":"Removes a useful sound fallback but not the boundary proof."},{"slug":"bounded_domain_exhaustive_search","disposition":"selected_supporting","contribution_type":"TEST_DESIGN","adaptation_or_rejection":"Exhaust synthetic finite model sets and fixed horizons.","counterfactual_removal":"The first test loses complete within-bound coverage."},{"slug":"computability_boundary_decision_record","disposition":"selected_supporting","contribution_type":"OPERATIONAL","adaptation_or_rejection":"Version the boundary, guarantee, assumptions, and triggers.","counterfactual_removal":"Guarantees can drift without provenance."},{"slug":"computational_complexity_analysis","disposition":"selected_supporting","contribution_type":"OPERATIONAL","adaptation_or_rejection":"Gate decidable fragments on time and memory growth.","counterfactual_removal":"Decidable may be mistaken for deployable."},{"slug":"constructive_algorithm_and_correctness_proof","disposition":"selected_supporting","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Witness total analyzers for claimed decidable fragments.","counterfactual_removal":"Positive decidability claims lack constructive support."},{"slug":"diagonalization_impossibility_proof","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Redundant because a domain-specific halting reduction is more reviewable.","counterfactual_removal":"No material change."},{"slug":"enumeration_and_dovetailing","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Unbounded operation is unsuitable; use bounded semi-decision.","counterfactual_removal":"No material change."},{"slug":"fallback_mode_router","disposition":"selected_load_bearing","contribution_type":"OPERATIONAL","adaptation_or_rejection":"Route by enforceable class to exact, abstract, bounded, or escalation modes.","counterfactual_removal":"Weaker outputs can be misread as universal verdicts."},{"slug":"halting_problem_reduction","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Attempt a checked embedding from halting to unbounded toxic-endpoint reachability.","counterfactual_removal":"The proposed impossibility boundary lacks decisive evidence."},{"slug":"language_fragment_restriction","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Enforce syntax admitting a total reachability procedure.","counterfactual_removal":"No guaranteed exact operational region remains."},{"slug":"many_one_reduction_proof","disposition":"selected_supporting","contribution_type":"TEST_DESIGN","adaptation_or_rejection":"Supply totality, computability, and answer-preservation obligations for the reduction.","counterfactual_removal":"Reduction rigor materially weakens."},{"slug":"promise_problem_restriction","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Unenforced biological promises could yield authoritative wrong answers; prefer checked syntax.","counterfactual_removal":"No material change."},{"slug":"proof_by_counterexample","disposition":"selected_supporting","contribution_type":"TEST_DESIGN","adaptation_or_rejection":"Use in-scope synthetic models to refute overbroad analyzer claims.","counterfactual_removal":"Cheap detection of universal overclaims is lost."},{"slug":"proof_checking","disposition":"selected_load_bearing","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"Independently verify the reduction and constructive proofs.","counterfactual_removal":"A proof gap could authorize a false boundary."},{"slug":"reduction_direction_checklist","disposition":"selected_supporting","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"Gate source-to-target direction and theorem assumptions.","counterfactual_removal":"A reversed or mismatched reduction may pass review."},{"slug":"semi_decision_with_explicit_unknown","disposition":"selected_load_bearing","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"Bound toxic-witness search and return UNKNOWN, never SAFE, on exhaustion.","counterfactual_removal":"Timeout can become a false negative."},{"slug":"theorem_prover_guided_search","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Optional proof discovery adds no necessary guarantee beyond independent checking.","counterfactual_removal":"No material change."},{"slug":"turing_reduction_analysis","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Oracle-relative degrees are unnecessary for the binary project boundary.","counterfactual_removal":"No material change."}],"causal_chain":["Make the universal toxicity-reachability claim and execution model explicit.","Test the unrestricted claim with a checked halting reduction while seeking total procedures for restricted fragments.","Enforce decidable fragment membership and route other inputs to sound abstraction, bounded witness search, or escalation.","Preserve UNKNOWN and attach the applicable guarantee to every output.","Recheck when model expressiveness or assumptions change, reducing false universal assurance and wasted solver work. [INFERENCE]"],"baseline":"Continue building one unrestricted Boolean analyzer; treat completed searches as evidence and timeouts as SAFE, failure, or ad hoc exceptions. [HYPOTHESIS]","nearest_rival":"Treat the issue solely as state-space explosion and optimize simulation or hardware without first separating computability from complexity.","authority_safety":{"affected_parties":["patients and trial participants","exposed populations","animals in toxicology studies","model users and safety reviewers"],"decision_authority":"A designated computational-toxicology model owner and independent safety-methods reviewer may approve analysis labels; clinical, regulatory, or exposure decisions remain with their lawful authorities.","authorized_first_step":"Run a sandboxed test on synthetic models only: formalize one expressive model language and one enforceable fragment, check the proposed reduction, exhaust a finite corpus within fixed bounds, and audit output labels.","excluded_actions":["No dosing, trial-enrollment, product-release, exposure-limit, or patient-care decision from pilot outputs.","No SAFE verdict from timeout, UNKNOWN, failed proof search, or out-of-scope input.","No unrestricted impossibility claim until the reduction and theorem assumptions pass independent review."],"halt_rollback":"Halt if an output is mislabelled, fragment membership is unenforceable, the reduction fails review, or a synthetic toxic witness is reported SAFE; disable decision use, preserve logs, revert to UNKNOWN/manual review, and correct the boundary record."}},"negative_tests":{"strongest_counterevidence":"The deployed model class may already be finite, bounded, and effectively enumerable with a known total reachability algorithm; then the issue is complexity, not computability. [HYPOTHESIS]","analogy_break":"Real organisms are not automatically arbitrary programs, and empirical uncertainty or biological complexity does not establish undecidability. The mapping applies only to the declared executable model language and formal reachability claim.","failure_condition":"The approach fails operationally if the restricted fragment excludes most decision-relevant models, abstractions generate unusable alarms, or users treat UNKNOWN as SAFE. [HYPOTHESIS]","problem_falsifier":"Show that no class-wide exact terminating claim exists, or that every accepted model is mechanically confined to a finite/decidable class with a valid total procedure.","intervention_falsifier":"After the bounded pilot, the boundary workflow does not improve correct differentiation of SAFE, TOXIC, UNKNOWN, timeout, and out-of-scope relative to baseline, or its checked labels still permit a synthetic false-SAFE result.","risks":["A formally correct model verdict may not represent biological reality.","Restriction may hide omitted mechanisms or relevant long-horizon toxicity.","False alarms may cause alert fatigue.","Boundary records may become stale or be used to legitimize a wrong formalization.","Downstream users may erase guarantee labels."]},"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":0.84,"generator_notes":"Closed-book structural transfer. The candidate concerns formal executable-model reachability, not a claim that biological toxicity itself is undecidable."}