{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp06_four_proposal_generalization60_20260803","source_assessment_id":"computability_boundary_mapping__human_computer_interaction:P3:v0","cell_id":"computability_boundary_mapping__human_computer_interaction","proposal_index":3,"qualification":"EMPIRICAL_PARTNER_CANDIDATE","criteria":{"specific_differentiated_claim":{"status":"YES","reason":"The falsifiable claim is that combining Wizard-of-Oz provenance, a nonvisual task-semantic contract, enforceable finite-profile checking, and explicit divergence/unknown labels reduces unsupported equivalence conclusions relative to scripted-task review and finite checking without provenance or unknown semantics."},"credible_problem_signal":{"status":"YES","reason":"Government sources establish consequential accessibility barriers and institutional modernization obligations, while HCI evidence establishes Wizard-of-Oz transparency, variability, and reproducibility risks. The exact migration overclaim's prevalence remains uncertain, but the external signal is credible rather than speculative problem fishing."},"identifiable_partner_or_adopter":{"status":"YES","reason":"A federal or comparable public-sector accessibility modernization program, represented by its product sponsor, Section 508 or accessibility lead, IT governance authority, and an independent formal-methods reviewer, is a concrete partner and adopter class."},"partner_access_is_necessary":{"status":"YES","reason":"A real program must provide its migration requirements, protected nonproduction models or fixtures, Wizard-of-Oz evidence practices, semantic-contract review, and sponsor-facing decision workflow; public research cannot determine whether the overclaim occurs, whether the contract covers material behavior, or whether UNKNOWN and provenance survive operational handoffs."},"safe_authorized_first_step":{"status":"YES","reason":"The authorized study is offline and nonproduction, leaves applications unchanged, assigns decisions to existing sponsor and accessibility authorities, prohibits migration and universal-equivalence claims, and has explicit halt and rollback conditions."},"bounded_decisive_empirical_design":{"status":"YES","reason":"The preregistered 12-pair study has three comparators, seeded ground truth, frozen encodings and bounds, measurable verdict and workflow outcomes, and explicit go/no-go falsifiers. It can decisively reject the integrated gate before deployment even though it cannot estimate population-wide prevalence."},"no_material_negative_gate":{"status":"YES","reason":"The external evaluation marks the problem gate UNCERTAIN rather than NO and marks the adopter, incremental-claim, bounded-step, safety/authority, and cost-scope gates YES. No unresolved safety, authority, collision, or established-practice gate is materially negative."},"not_merely_more_research":{"status":"YES","reason":"The next action is a specified partner-run fixture experiment with fixed cases, comparators, metrics, success conditions, and falsification rules, not an open-ended request for additional literature or market research."}},"uncertainty_types":["PROBLEM_PREVALENCE","ADOPTER_PULL","INCREMENTAL_EFFECT","WORKFLOW_FIT","DATA_ACCESS","COST_SCOPE"],"partner_profile":"One accessibility modernization program considering migration of a legacy interactive application, with an accountable product sponsor, accessibility or Section 508 lead, access to protected nonproduction artifacts, and an independent formal-methods reviewer.","required_access":"Authorized access to a frozen nonproduction legacy/replacement model pair or representative fixtures, actual migration and preservation requirements, a Wizard-of-Oz transcript and provenance workflow, accessibility review of the task-semantic contract, and the sponsor-facing reporting process; no production deployment or participant data is required.","bounded_empirical_test":"Run the preregistered 12-pair nonproduction study under a frozen task-semantic contract: three equivalent finite pairs with presentation differences, four finite pairs with seeded semantic divergences, two unrestricted divergent pairs, one nontermination-before-divergence pair, and two bounded no-divergence pairs. Compare scripted-task/screenshot review, finite checking without provenance or explicit unknown, and the full gate on verdict accuracy, false-equivalence rate, replayability, starvation, provenance retention, reviewer effort, and sponsor interpretation.","success_condition":"Finite-profile membership is enforceable; every seeded finite verdict is correct; every reported divergence replays; nontermination does not starve the later divergence; human-supplied behavior remains labeled; bounded non-finding remains UNKNOWN_EQUIVALENCE in sponsor-facing output; accessibility review finds no material omission in the frozen contract; and the full gate produces fewer unsupported equivalence interpretations than both comparators.","falsification_condition":"Reject or redesign the intervention if profile membership is bypassable, any seeded finite pair is misclassified, a reported divergence cannot be replayed, nontermination starves a later trace, human behavior is represented as generated, bounded non-finding becomes equivalence, or accessibility reviewers identify a material task commitment outside the observation contract.","rationale":"The candidate retains a distinct integration claim despite substantial adjacent prior art, has credible external problem and authority signals, and proposes a safe comparator-based experiment. Its decisive uncertainties concern real migration requirements, semantic fidelity, provenance retention, and sponsor interpretation, which require authorized access to an adopter workflow and cannot be resolved through ordinary public web research."}