{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp04_retrieval_first_paired20_20260802","cell_id":"computability_boundary_mapping__art_aesthetics","arm":"RETRIEVAL_FIRST","round_index":0,"hypotheses":[{"hypothesis_id":"H1","title":"Sound hazard screening for interactive installations","problem":"Pre-opening tests cannot establish that an arbitrary interactive artwork will never freeze, emit prohibited flashes, or enter another hazardous state.","affected_stakeholder":"Exhibition technicians and visitors vulnerable to visual hazards.","workflow_boundary":"Transfer from artist acceptance testing to public exhibition approval.","failure_mode":"Finite playtests are presented as universal safety evidence, and timeouts are treated as safe outcomes.","unit_of_analysis":"One executable installation build, its event model, and a formal hazard property.","causal_lever":"Replace concrete behavior with a finite over-approximation that certifies safety soundly while labeling possible violations as alarms or unknowns.","archetype_mapping":"Map the unrestricted behavior question, decide a finite abstraction, and route inconclusive cases without converting them into safe verdicts.","expected_value":"Could reduce undetected hazards while making false-alarm and residual-uncertainty costs explicit.","falsifiable_claim":"Across a retrospective set of at least 20 interactive builds, abstraction-based screening will identify at least one reachable hazard missed by recorded playtests and produce no certified-safe build with a reproduced in-model hazard.","diversity_rationale":"Targets runtime safety at exhibition acceptance; its unit is an executable build and its lever is sound over-approximation.","mechanism_slugs":["abstract_interpretation_or_model_checking","fallback_mode_router","computability_boundary_decision_record"],"search_questions":["Have museums or creative-coding platforms used formal verification for interactive-art safety?","Which installation hazards can be represented as finite-state properties?","What false-alarm rates make sound screening operationally unusable?"]},{"hypothesis_id":"H2","title":"Decidable composition-rule language for creative tools","problem":"A design tool promises an always-terminating checker for compositional rules embedded in unrestricted artist-authored scripts.","affected_stakeholder":"Creative-tool developers and artists who rely on authoring-time feedback.","workflow_boundary":"Rule authoring and linting before an artwork enters critique or production.","failure_mode":"An unrestricted scripting escape hatch defeats total analysis, but the interface still forces pass or fail.","unit_of_analysis":"One versioned rule-language grammar and its corpus of authored compositions.","causal_lever":"Restrict rules to a mechanically enforceable decidable fragment and supply a total, proved checker for that fragment.","archetype_mapping":"Trade expressive power for a declared solvability guarantee, then separately assess the checker's practical complexity.","expected_value":"Could provide dependable feedback for common rules while exposing the coverage lost through restriction.","falsifiable_claim":"In a corpus study, the fragment will express at least 75% of currently used machine-checkable composition rules, terminate on every conforming case, and reject every nonconforming rule before analysis.","diversity_rationale":"Targets authoring-language design rather than runtime behavior; its lever is syntactic restriction and constructive totality.","mechanism_slugs":["language_fragment_restriction","constructive_algorithm_and_correctness_proof","computational_complexity_analysis"],"search_questions":["Which composition rules are currently encoded in creative software?","What decidable visual-constraint languages already exist?","How much real artistic expressiveness would the proposed fragment exclude?"]},{"hypothesis_id":"H3","title":"Bounded equivalence contracts for software-art conservation","problem":"Conservators are asked to certify that a migrated or emulated interactive artwork behaves identically to the original under every future interaction.","affected_stakeholder":"Time-based-media conservators and acquisition committees.","workflow_boundary":"Acceptance of a migrated work into the preserved collection record.","failure_mode":"Passing a replay suite is called universal equivalence, or an imprecise undecidability analogy is used to abandon comparison entirely.","unit_of_analysis":"A pair of artwork implementations plus a declared grammar and depth bound for audience interactions.","causal_lever":"Classify unrestricted equivalence separately, then publish a promise-bounded trace contract with exhaustive checking and explicit out-of-scope behavior.","archetype_mapping":"Test whether universal equivalence is undecidable under the actual model, while recovering a complete guarantee over an enforceable finite interaction region.","expected_value":"Could give conservators a defensible preservation claim without implying identity beyond observed or bounded behavior.","falsifiable_claim":"For at least 10 migration cases, the bounded contract will either expose a divergence absent from the existing replay suite or produce a complete certificate for the declared trace space; independent review will find no scope overstatement.","diversity_rationale":"Targets preservation equivalence at migration acceptance; its unit is an implementation pair and its lever combines impossibility classification with a bounded promise.","mechanism_slugs":["halting_problem_reduction","promise_problem_restriction","bounded_domain_exhaustive_search","proof_checking"],"search_questions":["How do conservators currently define behavioral authenticity for software art?","Are unrestricted equivalence claims actually made in migration practice?","Which interaction bounds would be meaningful and enforceable rather than convenient?"]},{"hypothesis_id":"H4","title":"Complete pre-mint audit of finite generative editions","problem":"Generative-art publishers sample outputs even when every token and render profile form a finite auditable edition.","affected_stakeholder":"Generative artists, publishers, and collectors relying on edition-level claims.","workflow_boundary":"Final code freeze before minting or releasing a numbered edition.","failure_mode":"Rare trait collisions or prohibited outputs escape sampling because a large finite search is confused with undecidability.","unit_of_analysis":"One frozen generator crossed with all declared token identifiers and supported render profiles.","causal_lever":"Enumerate the complete finite domain, retain coverage certificates and counterexamples, and use complexity analysis to reject infeasible bounds before promising completion.","archetype_mapping":"Place the edition on the decidable side by explicit finiteness, without generalizing the certificate beyond the frozen encoding and bounds.","expected_value":"Could turn edition-wide aesthetic or contractual claims into checkable evidence and reveal when complete auditing is merely too costly.","falsifiable_claim":"For editions of at most 10,000 identifiers under a fixed render profile, exhaustive audit will finish within the publisher's pre-release compute budget and detect at least as many rule violations as stratified sampling, with complete domain coverage verifiable from the certificate.","diversity_rationale":"Targets finite release auditing; unlike the other hypotheses, the core issue is tractability versus decidability and the unit is an entire edition.","mechanism_slugs":["bounded_domain_exhaustive_search","computational_complexity_analysis","proof_by_counterexample"],"search_questions":["Do generative-art publishers make edition-wide trait or uniqueness guarantees?","What edition sizes and rendering costs make exhaustive audits feasible?","Which coverage certificates would collectors or auditors accept?"]},{"hypothesis_id":"H5","title":"Witness-or-unknown search for formal art briefs","problem":"Automated composition systems interpret failure to find an artwork satisfying a formal brief before a deadline as proof that no satisfying work exists.","affected_stakeholder":"Artists and commissioners using constraint-based or program-synthesis tools.","workflow_boundary":"Candidate generation between brief formalization and human selection.","failure_mode":"A stalled search becomes a false rejection, while one nonterminating candidate can starve other viable generators.","unit_of_analysis":"One machine-checkable brief, candidate encoding, witness checker, and search budget.","causal_lever":"Fairly dovetail candidate generators, return YES only with a checked witness, and return UNKNOWN—not NO—when the declared budget expires.","archetype_mapping":"Expose the task as one-sided recognition and make nontermination an explicit workflow state with governed escalation.","expected_value":"Could recover valid candidates hidden by serial search while preventing deadlines from masquerading as impossibility proofs.","falsifiable_claim":"On a preregistered benchmark containing planted satisfying witnesses and unresolved briefs, dovetailed search will return zero false NO verdicts and find more distinct planted witnesses within the same aggregate compute budget than the incumbent serial schedule.","diversity_rationale":"Targets synthesis scheduling and commissioning decisions; its unit is a formal brief and its lever is fair one-sided search with explicit unknown.","mechanism_slugs":["enumeration_and_dovetailing","semi_decision_with_explicit_unknown","theorem_prover_guided_search"],"search_questions":["Where are formal constraints used to generate or commission visual art?","Do current systems report timeout, failure, and unsatisfiability distinctly?","Which generator schedules and witness checkers permit fair comparison?"]}]}