{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp04_retrieval_first_paired20_20260802","cell_id":"computability_boundary_mapping__ethnography_qualitative_methods","arm":"PROPOSAL_FIRST","candidate_id":"qualitative_workflow_termination_boundary","hypothesis_id":null,"version":0,"title":"Computability-Bounded Assurance for Executable Qualitative Research Workflows","problem":"A research platform accepts user-authored executable workflows that code transcripts, revise codebooks, select subsequent cases, search for negative cases, and declare saturation. Its proposed validator is required to decide correctly and terminate for every workflow and every admissible transcript sequence, answering whether the workflow will eventually stop and whether it can stop before its declared evidentiary conditions are satisfied. Because the workflow language permits arbitrary computation and the transcript source is interactive, tests, timeouts, and successful pilots cannot establish this universal guarantee; conversely, an informal halting-problem analogy cannot establish impossibility without a model-matched reduction.","actors":["qualitative researchers","research participants whose accounts may be included or omitted","research-methods reviewers","institutional ethics and governance reviewers","research-platform engineers","downstream readers of qualitative findings"],"observable_state":"Submitted workflows sometimes run indefinitely, reach different stopping states under different transcript continuations, or time out during validation; nevertheless, the platform's interface offers only pass or fail and its documentation describes validation as universal.","consequence":"A timeout or incomplete search can be presented as failure, while an inadequately checked workflow can be presented as certified and terminate data collection prematurely. Either outcome can distort case selection, omit discordant accounts, delay studies, and give downstream users a stronger methodological assurance than the validator established.","affected_objective":"Methodologically faithful, auditable decisions about when an executable qualitative inquiry may stop without collapsing false, unknown, timeout, and out-of-scope states.","intervention":"Replace the universal Boolean-validation requirement with a model-relative boundary assessment. Precisely encode the workflow class, transcript interaction model, stopping property, and quantifiers; attempt a checked reduction showing whether a total exact validator for unrestricted executable workflows would decide program halting; and leave the status unresolved if the translation or semantic correspondence fails review. For operation, admit an enforceable finite-state workflow language with explicit stopping predicates, use exhaustive checking or sound over-approximation within its declared bounds, and route unrestricted workflows to UNKNOWN plus documented human methodological review. Version the shipped guarantees, assumptions, and recheck triggers without treating human judgment as an exact oracle.","structural_mapping":[{"archetype_element":"Problem-class specification","domain_realization":"The class is every syntactically valid executable coding, memoing, theoretical-sampling, negative-case, and stopping workflow accepted by the platform, paired with every transcript-event sequence allowed by the stated environment model."},{"archetype_element":"Instance representation and computation model","domain_realization":"An instance contains workflow source, initial codebook and memos, a formal stopping predicate, and an encoded transcript-event source; the contract states whether scripts are Turing-complete and whether future observations are fixed input, nondeterministic events, or external interaction."},{"archetype_element":"Universal guarantee and quantifier map","domain_realization":"The disputed requirement is one uniform validator that always terminates and correctly classifies termination and premature-stop safety for every admitted workflow and every admissible event continuation."},{"archetype_element":"Constructive and impossibility evidence","domain_realization":"A candidate decider requires totality and correctness arguments; a candidate impossibility result requires a computable translation from a known source problem whose answer is preserved under the exact workflow semantics."},{"archetype_element":"Decidable region","domain_realization":"A mechanically enforced finite-state workflow DSL, finite code and memo domains, and explicitly bounded transcript or transition depth form subclasses that can be exhaustively explored."},{"archetype_element":"Weaker honest fallback","domain_realization":"Sound finite abstraction may certify that no modeled path stops prematurely, while alarms remain inconclusive; bounded exploration reports only its bound, and unrestricted cases return UNKNOWN for methodological review."},{"archetype_element":"Status and traceability","domain_realization":"Each result records the workflow-language version, environment assumptions, property checked, mode used, bound or abstraction, guarantee label, unresolved residue, and triggers requiring reclassification."}],"mechanism_mapping":[{"mechanism_slug":"halting_problem_reduction","role":"Tests whether unrestricted workflow termination validation is undecidable by translating a program and input into a workflow that emits its declared terminal state and halts exactly when that program halts.","counterfactual_removal":"Without the reduction, the project would have only failed implementations and analogy, neither of which justifies abandoning a universal validator."},{"mechanism_slug":"reduction_direction_checklist","role":"Checks that the construction runs from the independently established source problem into workflow validation and names the encoding, quantifiers, and semantic assumptions.","counterfactual_removal":"A reversed or assumption-mismatched reduction could falsely label the platform requirement impossible."},{"mechanism_slug":"language_fragment_restriction","role":"Defines an enforceable finite-state qualitative-workflow DSL whose membership is syntactically decidable and whose loss of expressiveness is explicit.","counterfactual_removal":"The platform would lack a stable region in which an exact terminating assurance can legitimately be offered."},{"mechanism_slug":"bounded_domain_exhaustive_search","role":"Enumerates every transition and transcript-event choice within a declared finite configuration and depth, producing a complete result only for that bounded instance space.","counterfactual_removal":"Bounded workflows would be assessed by samples or heuristics despite admitting a complete, checkable analysis."},{"mechanism_slug":"abstract_interpretation_or_model_checking","role":"Over-approximates permitted workflow states so a successful premature-stop safety verdict is sound for the modeled system, while possible violations remain alarms requiring refinement or review.","counterfactual_removal":"The only operational options would be very small exhaustive bounds or unsupported exact claims for larger workflows."},{"mechanism_slug":"fallback_mode_router","role":"Routes each submission to exact finite checking, sound abstraction, bounded exploration, or UNKNOWN escalation and attaches the corresponding guarantee label.","counterfactual_removal":"Weaker outputs could be silently consumed as universal pass-or-fail certifications."},{"mechanism_slug":"computability_boundary_decision_record","role":"Versions the selected boundary, shipped guarantees, supporting arguments, unresolved assumptions, and recheck triggers such as added scripting constructs or changed transcript interaction.","counterfactual_removal":"A later language or environment change could silently invalidate the original boundary while the certification language remained unchanged."}],"causal_chain":["The platform admits expressive, user-authored iterative analysis and stopping logic.","Governance requests a correct, terminating validator over every workflow and admissible transcript continuation.","Finite tests and timeouts observe individual executions but do not discharge the universal termination and correctness quantifiers.","A model-matched halting reduction, if independently checked, places the unrestricted total-exact requirement beyond a decider; a failed reduction leaves that classification unresolved rather than rhetorically impossible.","An enforceable finite-state fragment and explicit finite bounds convert selected subclasses into decidable instance spaces.","Sound abstraction extends coverage while preserving only its one-sided guarantee; inconclusive alarms, timeouts, and out-of-fragment inputs remain UNKNOWN or OUT_OF_SCOPE.","A router labels each result with its actual guarantee, and a versioned decision record forces reconsideration when the workflow language, environment model, or stopping property changes.","Researchers and reviewers can then distinguish exact certification from bounded evidence, conservative warning, and human methodological judgment before authorizing data-collection closure."],"baseline":"Methodologists manually inspect submitted protocols, engineers run representative workflows in a sandbox, and a fixed timeout or reviewer judgment is converted into pass or fail. The baseline can identify concrete defects but supplies no class-wide termination proof and does not consistently preserve an unknown state.","nearest_rivals":["A sandbox with resource limits that kills long-running workflows; it guarantees operational termination of the sandbox but does not decide whether the workflow would eventually satisfy its stopping condition.","A heuristic saturation score based on recent code recurrence; it may support researcher judgment but cannot inherit an exact universal assurance.","Manual review using memoing, negative-case analysis, and protocol checklists; it can assess substantive methodological fitness but is not a declared computational oracle and may vary by reviewer.","A purely syntactic linter that rejects forbidden constructs; it can enforce local rules but does not by itself establish the semantic stopping and safety properties."],"remaining_contrastive_claim":"Relative to these rivals, the candidate's remaining claim is only that it makes the problem class and computational model explicit, demands checkable evidence before assigning a computability status, and structurally prevents bounded, one-sided, timed-out, or human-reviewed results from being labeled as universal exact certification.","authority_safety":{"decision_authority":"The study's principal investigator retains authority to continue or stop data collection, subject to applicable ethics and governance oversight; the platform validator may label guarantees and block unsupported automated certification but may not declare substantive saturation on the investigator's behalf.","authorized_first_step":"A methods reviewer and engineer may formalize one de-identified toy workflow family, its stopping predicate, and its event model; draft the candidate reduction; and test a finite-fragment checker without deploying it or changing an active study.","excluded_actions":["Automatically stopping recruitment or fieldwork","Overriding participant-consent, withdrawal, or ethics requirements","Treating UNKNOWN, timeout, or an abstraction alarm as NO or as methodological failure","Treating human approval as proof of a total computational guarantee","Uploading identifiable transcripts into the prototype","Generalizing a bounded result beyond its declared language, state space, or depth"],"halt_rollback":"Stop the assessment if the formal workflow no longer represents the actual analytic practice, the reduction cannot preserve the stated property, identifiable data enter the prototype, or users cannot distinguish guarantee labels. Withdraw any provisional certification, revert to documented manual review, retain the discrepancy as unresolved residue, and require a revised model before resuming."},"negative_tests":{"strongest_counterevidence":"A uniform validator for the exact declared unrestricted workflow and interaction model, accompanied by independently checkable proofs that it terminates and returns the correct stopping-safety classification on every admitted instance, would defeat the proposed impossibility boundary.","problem_falsifier":"The problem is absent if the platform accepts only a finite, mechanically enforced workflow language over a finite event space, makes no unrestricted class-wide claim, and already distinguishes exact, bounded, unknown, timeout, and out-of-scope results.","intervention_falsifier":"The intervention fails if its formalization omits meaning-determining workflow behavior, if the reduction does not preserve answers under the real execution semantics, if the supposedly decidable fragment is not enforceable, or if empirical users still interpret weaker guarantee labels as substantive saturation certification.","risks":["Formal verification may certify a faithful execution model but the wrong qualitative concept of saturation.","A restricted DSL may exclude reflexive, emergent, or context-sensitive analytic work and encourage covert escape hatches.","A coarse over-approximation may generate enough false alarms that reviewers ignore it.","The decision record may institutionalize an incorrect boundary if independent review is weak or triggers become stale.","Routing to human review may concentrate unmodeled authority, delay studies, or conceal inconsistent judgments.","Workflow traces and counterexamples may expose sensitive participant material unless de-identified and access-controlled."]},"next_evidence_step":"Within a two-week, non-deployment exercise, specify six synthetic workflows in one tiny formal language: two terminating safely, one stopping before its declared negative-case check, one looping, one whose termination depends on an encoded simulated program, and one outside the finite fragment. Have one engineer draft the source-to-workflow reduction and a separate reviewer apply the direction checklist; exhaustively check only a predeclared finite configuration and depth; record exact, bounded, UNKNOWN, and OUT_OF_SCOPE outputs. The step supports only a decision about whether the formalization and proposed boundary assessment deserve a larger study.","prior_art_status":"UNSEARCHED","revision_record":{"parent_version":null,"progress_targets_addressed":[],"conceptual_changes":[],"operational_changes":[],"evidence_changes":[],"claim_changes":[]}}