{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp03_full320_20260801","cell_id":"computability_boundary_mapping__psychology","trajectory_id":"R","attempt_index":0,"archetype_slug":"computability_boundary_mapping","domain_slug":"psychology","decision":"CANDIDATE","problem_id":"universal_executable_psychological_model_behavior_analyzer_overclaim","causal_lever_id":"model_relative_behavior_reachability_boundary_with_labeled_fallbacks","proposal":{"problem":"Psychological-modeling projects may promise an exact, always-terminating analyzer that decides whether any arbitrary executable cognitive, behavioral, or therapy-process model will ever produce a specified state, such as harmful action or symptom remission. When the model language is open-ended, teams may treat timeout as absence, generalize bounded simulation, or invoke undecidability without proving that the language and property satisfy the theorem. This can misstate model safety and waste effort on an impossible universal guarantee.","actors_substrate":["computational psychology researchers","executable-model and digital-intervention developers","independent methods and safety reviewers","clinicians or policy users consuming model verdicts","people represented by or exposed to model-guided decisions"],"observable_state":"Analyzers time out or return forced Boolean verdicts; bounded simulations are advertised universally; model-language restrictions remain undocumented; and reviewers cannot distinguish false, unknown, out-of-scope, nontermination, or model failure.","consequence":"Potentially harmful modeled trajectories may be falsely cleared, useful models may be rejected, and research or safety resources may be committed to an unattainable universal analyzer.","affected_objective":"Truthful, auditable assessment of executable psychological models without converting formal uncertainty into clinical certainty.","structural_mapping":[{"archetype_element":"unbounded problem class","domain_realization":"Arbitrary executable psychological models paired with reachability questions about a defined modeled behavior or state.","claim_kind":"HYPOTHESIS"},{"archetype_element":"total exact terminating requirement","domain_realization":"A requested YES/NO answer for every well-formed model-property pair, with guaranteed termination and no abstention.","claim_kind":"HYPOTHESIS"},{"archetype_element":"computability-model boundary","domain_realization":"The result depends on the model language, encoding, permitted simulation or oracle access, and whether the target property concerns formal traces rather than real people.","claim_kind":"INFERENCE"},{"archetype_element":"weaker honest fallback","domain_realization":"Enforceable decidable fragments, bounded exploration, or sound one-sided analysis returning explicit UNKNOWN outside its guarantee.","claim_kind":"HYPOTHESIS"},{"archetype_element":"guarantee traceability","domain_realization":"Each verdict carries its model version, scope, method, bound, assumptions, and recheck triggers.","claim_kind":"HYPOTHESIS"}],"component_map":[{"component":"Problem-Class Specification","status":"adapted","domain_realization":"Define model-property pairs and whether the question is modeled-state reachability, inevitability, or nonoccurrence."},{"component":"Instance Representation Contract","status":"adapted","domain_realization":"Fix model syntax, initial conditions, environmental inputs, and behavioral-state encoding."},{"component":"Computation Model Contract","status":"direct","domain_realization":"Declare execution semantics, resources, interaction, randomness treatment, and any human or data oracle."},{"component":"Solvability Guarantee Profile","status":"direct","domain_realization":"Specify exact decision, one-sided recognition, sound abstraction, bounded result, or unresolved status."},{"component":"Quantifier and Scope Map","status":"direct","domain_realization":"Separate every model and input from one model, one population, one horizon, or one parameter range."},{"component":"Computability Status Lattice","status":"direct","domain_realization":"Classify each formal question as decidable, recognizable, relative, undecidable, or unresolved."},{"component":"Constructive Procedure Witness","status":"adapted","domain_realization":"For a decidable fragment, supply the analyzer with termination and correctness arguments."},{"component":"Reduction Preservation Contract","status":"direct","domain_realization":"Require a total computable translation preserving the target reachability answer."},{"component":"Computability Impossibility Certificate","status":"direct","domain_realization":"Accept impossibility only from a checked reduction or direct proof matching the model language."},{"component":"Assumption Register","status":"direct","domain_realization":"Record expressiveness, semantics, input admissibility, target-property definition, and oracle assumptions."},{"component":"Decidable Subclass Map","status":"adapted","domain_realization":"Identify enforceable finite-state, bounded-horizon, or syntactically restricted model classes."},{"component":"One-Sided Recognition Contract","status":"adapted","domain_realization":"Permit confirmed reachable states with witnesses while withholding unproved nonreachability."},{"component":"Unknown and Nontermination Policy","status":"direct","domain_realization":"Keep UNKNOWN, timeout, out-of-scope, and execution failure distinct from NO."},{"component":"Fallback Solution Contract","status":"direct","domain_realization":"State the exact weaker guarantee used when total analysis is unavailable."},{"component":"Computability Guarantee Record","status":"direct","domain_realization":"Version-link each formal verdict to proof, analyzer, assumptions, and scope."},{"component":"Recheck Trigger","status":"direct","domain_realization":"Reclassify after language, semantics, target property, bound, or oracle changes."},{"component":"Termination Condition","status":"direct","domain_realization":"Define proof-based termination or a resource bound ending in UNKNOWN."},{"component":"Scope Boundary","status":"direct","domain_realization":"Mechanically reject or quarantine model-property pairs outside the supported class."},{"component":"Decision Record","status":"direct","domain_realization":"Record why a specific guarantee and fallback were selected."},{"component":"Uncertainty Residue","status":"adapted","domain_realization":"List unresolved proof obligations and the gap between modeled and human behavior."},{"component":"Independent Proof Review","status":"direct","domain_realization":"Have a reviewer check theorem statement, formalization, reduction, and assumptions."},{"component":"Complexity Follow-On Gate","status":"direct","domain_realization":"Only after decidability, assess whether analysis is feasible at relevant model sizes."}],"mechanism_dispositions":[{"slug":"abstract_interpretation_or_model_checking","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Over-approximate modeled behaviors so a proved absence of harmful abstract traces can support a one-directional verdict; otherwise report possible violation.","counterfactual_removal":"Removing it eliminates the principal sound fallback for models too expressive for exact analysis."},{"slug":"bounded_domain_exhaustive_search","disposition":"selected_supporting","contribution_type":"TEST_DESIGN","adaptation_or_rejection":"Exhaustively check a declared finite horizon or state bound and label results as bounded.","counterfactual_removal":"The pilot loses a complete within-bound benchmark, but the boundary classification remains possible."},{"slug":"computability_boundary_decision_record","disposition":"selected_supporting","contribution_type":"OPERATIONAL","adaptation_or_rejection":"Version the chosen boundary, guarantee, assumptions, and recheck triggers.","counterfactual_removal":"The mathematical result remains, but institutional scope drift becomes harder to detect."},{"slug":"computational_complexity_analysis","disposition":"selected_supporting","contribution_type":"OPERATIONAL","adaptation_or_rejection":"Assess scaling only after a fragment is classified decidable.","counterfactual_removal":"Correct computability claims remain, but deployability cannot be responsibly judged."},{"slug":"constructive_algorithm_and_correctness_proof","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Require an actual total correct analyzer before calling a fragment decidable.","counterfactual_removal":"Positive classifications would lack their strongest constructive witness."},{"slug":"diagonalization_impossibility_proof","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"A direct self-reference proof is unnecessary unless it fits the selected model language better than reduction.","counterfactual_removal":"No change because the proposed impossibility route uses a checked reduction."},{"slug":"enumeration_and_dovetailing","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Unbounded dovetailing conflicts with the operational requirement to terminate; its honest one-sided logic is retained through bounded semi-decision.","counterfactual_removal":"No material change because the bounded protocol supplies the deployed behavior."},{"slug":"fallback_mode_router","disposition":"selected_load_bearing","contribution_type":"OPERATIONAL","adaptation_or_rejection":"Route in-scope queries to exact, abstract, bounded, or escalation modes and attach the applicable guarantee.","counterfactual_removal":"Weaker methods could be applied without reliable scope or guarantee labels."},{"slug":"halting_problem_reduction","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Test whether a total computable embedding from halting instances into unrestricted model reachability can be constructed.","counterfactual_removal":"The conjectured universal impossibility would lack a decisive certificate."},{"slug":"language_fragment_restriction","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Define mechanically enforceable model fragments admitting total reachability analysis.","counterfactual_removal":"The intervention would diagnose a boundary without recovering a guaranteed useful region."},{"slug":"many_one_reduction_proof","disposition":"selected_supporting","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"Use its totality, computability, and answer-preservation obligations to formalize the halting reduction.","counterfactual_removal":"The reduction would be more vulnerable to an invalid or one-way mapping."},{"slug":"promise_problem_restriction","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"An unenforced promise permits authoritative answers on violations; prefer mechanically checked syntax or explicit out-of-scope routing.","counterfactual_removal":"No loss because enforceable fragment membership supplies a safer boundary."},{"slug":"proof_by_counterexample","disposition":"selected_supporting","contribution_type":"TEST_DESIGN","adaptation_or_rejection":"Use one in-scope failure to refute an advertised universal analyzer, without treating failed search as proof.","counterfactual_removal":"Universal overclaims become harder to falsify cheaply, though status proofs remain available."},{"slug":"proof_checking","disposition":"selected_load_bearing","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"Independently check constructive and impossibility certificates and expose premises.","counterfactual_removal":"Boundary decisions would depend excessively on author authority and hidden assumptions."},{"slug":"reduction_direction_checklist","disposition":"selected_supporting","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"Gate the impossibility verdict on source-to-target direction and explicit assumptions.","counterfactual_removal":"A reversed or incomplete reduction could be mistaken for proof."},{"slug":"semi_decision_with_explicit_unknown","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Return witnessed YES before a resource bound and UNKNOWN at the bound, never fabricated NO.","counterfactual_removal":"Timeout would again be vulnerable to interpretation as behavioral nonoccurrence."},{"slug":"theorem_prover_guided_search","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Optional automation cannot replace correct formalization or independent checking in the first test.","counterfactual_removal":"No hard-gate changes; proofs may be developed and checked without automated discovery."},{"slug":"turing_reduction_analysis","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Relative degrees and adaptive oracle access are unnecessary for the initial binary boundary claim and obscure one-sided structure.","counterfactual_removal":"No material change to the proposed many-one impossibility test."}],"causal_chain":["Make the executable model class, property, semantics, quantifiers, and guarantee explicit.","Attempt constructive total analysis and a source-to-target impossibility reduction under the same contracts.","Independently check the successful certificate and leave mismatched or incomplete cases unresolved.","Enforce decidable fragments and route remaining queries to sound abstract or bounded methods.","Preserve UNKNOWN and attach scope and guarantee labels, reducing false clearance and overclaiming."],"baseline":"Teams run simulations or heuristic analyzers until a timeout, then report a Boolean conclusion or generalize success from tested models without a class-wide termination argument.","nearest_rival":"Treat the issue solely as computational-complexity or engineering capacity and optimize simulation, which helps only after in-principle decidability and scope are established.","authority_safety":{"affected_parties":["people represented by psychological models","patients or users exposed to model-guided interventions","clinicians and researchers relying on verdicts","model developers and reviewers"],"decision_authority":"Model owners and an independent formal-methods reviewer may approve research guarantees; clinical or human-facing use additionally remains under the relevant clinical, ethics, and institutional authority.","authorized_first_step":"Offline pilot on one declared executable model language: specify one reachability property, audit expressiveness, attempt both a total analyzer for a restricted fragment and a checked halting reduction for the unrestricted class, then compare exact, abstract, and bounded outputs on synthetic models.","excluded_actions":["changing patient care from the pilot","labeling a person dangerous or safe from formal-model reachability","withholding support based on UNKNOWN","claiming real human behavior is computable or undecidable from a theorem about its model"],"halt_rollback":"Stop the pilot if the behavioral property cannot be formalized without material semantic loss, the reduction or constructive proof fails independent review, the fragment boundary cannot be enforced, or UNKNOWN is consumed as NO. Withdraw the affected guarantee and revert to explicitly exploratory simulation."}},"negative_tests":{"strongest_counterevidence":"Many deployed psychological models may already be finite, bounded, or statistically specified, making their formal analysis decidable; their important failures may instead arise from measurement error, construct validity, distribution shift, or poor prediction of people.","analogy_break":"An executable psychological model is not a person. A computability theorem can classify a formal model-property pair but cannot establish that real behavior is algorithmically predictable, unreachable, safe, or impossible.","failure_condition":"The intervention fails if its formal boundaries are ignored downstream, the abstraction drops real modeled behaviors, the restricted fragment omits practically essential cases, or UNKNOWN rates make the fallback unusable.","problem_falsifier":"The problem is falsified if scoped projects make no universal exact terminating claim, all relevant model classes are demonstrably finite or otherwise decidable, and observed errors concern only empirical validity or resource scaling rather than collapsed computability statuses.","intervention_falsifier":"The proposed lever is falsified if a contract-matched total correct analyzer already covers the declared unrestricted class, or a bounded comparison shows that boundary routing and explicit UNKNOWN do not reduce false Boolean conclusions while retaining useful coverage.","risks":["Formal precision may create false confidence about correspondence to human behavior.","A coarse abstraction may produce excessive false alarms.","A purportedly safe fragment may be bypassed through extensions or external services.","UNKNOWN may be suppressed by clinical or product interfaces.","Impossibility language may discourage solvable bounded or empirical work."]},"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":"Candidate fit depends on the target being an executable formal psychological-model language with a class-wide guarantee. The packet provides no evidence that a particular deployed language is expressive enough for the proposed reduction, so undecidability and operational prevalence remain hypotheses."}