{"schema_version":1,"assessment_id":"eoa_inverse_innovation_exp03_opportunity320_20260801","source_experiment_id":"eoa_inverse_innovation_exp03_full320_20260801","cell_id":"negative_space_design__tech_ethics_ai_governance","archetype_slug":"negative_space_design","domain_slug":"tech_ethics_ai_governance","title":"Protected negative-space format for AI deployment review","opportunity_summary":"Test a review format that isolates unresolved hazards, labels why evidence is absent, preserves immediate retrieval of omitted material, and adds a chair-held interpretive interval before discussion and decision. The proposal targets missed hazards, misread evidence gaps, and suppressed dissent without changing production decisions during the first test.","adopter_authorizer":"An AI governance chair can authorize the proposed shadow-format simulation; only the established approval body can authorize changes to deployment decisions or governance policy.","scores":{"meaningful_impact":{"score":4,"rationale":"If the specified failure occurs, missed high-severity hazards and premature approval could materially affect reviewers, deployment operators, and people exposed to the AI system. The consequence is important, although the packet does not establish how often it occurs."},"stakeholder_pull":{"score":3,"rationale":"The proposal names governance reviewers, chairs, model owners, risk staff, and affected people with a shared interest in accurate and contestable decisions. It supplies no evidence of expressed demand, baseline dissatisfaction, or willingness to adopt the format."},"incremental_advantage":{"score":3,"rationale":"Relative to the stated risk-ranking dashboard, the composition additionally protects an interpretive interval, creates space for dissent, labels absent evidence, and requires recoverability. Whether those additions improve decisions rather than hide dependencies remains untested."},"distinctiveness_plausibility":{"score":2,"rationale":"The combined use of reversible subtraction, labeled absence, protected silence, and retrieval is specifically articulated, but prior art is explicitly unsearched. The packet therefore cannot support novelty or uncommonness beyond its stated comparison with one nearest rival."},"technical_implementability":{"score":4,"rationale":"A non-production crossover using synthetic or de-identified cases, alternate packet formats, timed retrieval, and recorded outcomes is technically straightforward. Secure retrieval, accessibility, relevance judgments, and reliable coding of dissent still require careful implementation."},"adoption_authority_feasibility":{"score":4,"rationale":"The packet identifies both the chair authorized to approve a shadow-format test and the established body retaining deployment and policy authority. Broader adoption would still require that body's acceptance and coordination with compliance, accessibility, and model-owner stakeholders."},"evidence_readiness":{"score":4,"rationale":"The candidate specifies a conventional-format comparison, observable outcomes, separate problem and intervention falsifiers, prohibited actions, and halt criteria. Decision thresholds, case sampling, panel recruitment, and measurement reliability remain to be defined."},"safety_net_benefit":{"score":4,"rationale":"The format could provide an additional check against premature approval by making unresolved hazards, evidence absence, and dissent more salient. The proposed first step preserves the full packet, permits immediate rollback, and cannot control production deployment, although omission and silence can themselves create risk."},"scalability":{"score":3,"rationale":"The format is a repeatable review-process intervention rather than specialized physical infrastructure, suggesting possible transfer across panels. Scaling depends on local review authority, case-specific relevance judgments, retrieval systems, accessibility, organizational power dynamics, and facilitator discipline."}},"score_confidence":"MODERATE","costs":{"first_evidence":{"band_2026_usd":"10K_TO_50K","scope":"Design and run one bounded crossover simulation, including case preparation, conventional and negative-space packet production, panel time, facilitation, accessibility checks, outcome coding, and analysis.","confidence":"MODERATE","assumptions":["Synthetic or already-authorized de-identified historical cases are available without new data acquisition.","The test uses existing review personnel and ordinary meeting and document systems.","The scope excludes production deployment decisions and major software development."]},"initial_deployment_startup":{"band_2026_usd":"50K_TO_250K","scope":"Prepare the format for controlled organizational use after favorable evidence, including governance approval, workflow design, secure retrieval controls, absence-label taxonomy, accessibility review, training, compliance review, and evaluation instrumentation.","confidence":"LOW","assumptions":["Deployment is limited to one organization or governance function.","Existing document, identity, and audit systems can be adapted rather than replaced.","Legal and policy changes are limited, but cross-functional staff coordination is required."]},"operational_launch":{"band_2026_usd":"50K_TO_250K","scope":"Launch across a limited set of AI review panels with facilitated sessions, support for packet conversion and retrieval, adverse-effect monitoring, quality assurance, and comparison against the conventional process.","confidence":"LOW","assumptions":["Launch remains reversible and runs alongside established approval controls.","No custom enterprise platform or large-scale historical-data remediation is required.","Panels require active support while relevance rules and silence protocols are calibrated."]},"annual_recurring":{"band_2026_usd":"10K_TO_50K","scope":"Maintain templates and retrieval controls, train new participants, audit omissions and absence labels, monitor accessibility and power-related effects, and periodically reassess outcomes for one governance function.","confidence":"LOW","assumptions":["The organization embeds most review work in existing staff roles.","Review volume is moderate and the underlying document systems remain available.","Material expansion across business units, jurisdictions, or high-volume review programs would raise the band."]}},"research_burden":"MODERATE","earliest_credible_horizon":"3_TO_12_MONTHS","pipeline_gates":{"recognizable_externally_supportable_problem":{"status":"YES","reason":"The packet defines an observable problem through missed unresolved hazards, incorrect interpretation of blank evidence fields, limited dissent, packet crowding, and chair speaking time, with a baseline falsifier. Its prevalence is not established."},"identifiable_adopter_or_authorizer":{"status":"YES","reason":"The AI governance chair is identified as the authorizer for a shadow-format test, while the established approval body retains authority over policy and deployment decisions."},"distinct_testable_incremental_claim":{"status":"YES","reason":"The candidate claims that reversible evidence isolation, labeled absence, protected interpretation, and chair-held silence improve hazard detection, absence interpretation, and dissent relative to the conventional packet and stated ranking-dashboard rival."},"bounded_next_evidence_step":{"status":"YES","reason":"The sealed candidate authorizes a non-production crossover simulation using synthetic or de-identified cases, a conventional comparison, specified outcome classes, and explicit adverse-effect falsifiers."},"no_unresolved_safety_or_authority_stop":{"status":"YES","reason":"The first step cannot authorize deployment, retains the full packet for retrieval, prohibits hiding required evidence and coercing speech, and specifies immediate halt and rollback conditions. Known omission, accessibility, distress, and power risks must still be measured."},"implementation_cost_scope_and_range":{"status":"YES","reason":"A bounded first test and limited organizational implementation can be scoped using broad resource-equivalent bands, with stated assumptions. Exact cost and the cost of wider scaling remain uncertain."}},"blocking_evidence":["Baseline observations showing whether reviewers actually miss or misclassify unresolved high-severity risks, misread evidence gaps, or withhold dissent under the conventional process.","Crossover evidence that the negative-space version improves hazard detection, absence interpretation, or dissent relative to the conventional version without increasing missed context, retrieval delay, distress, or overconfident approval.","Evidence that omitted material remains complete, auditable, accessible, and immediately retrievable across relevant reviewer needs.","Assessment of whether protected silence creates coercion, disapproval cues, or unequal participation for lower-power reviewers.","Prior-art comparison establishing whether the combined design differs materially from existing governance review formats before any distinctiveness claim.","Adopter evidence that the chair, established approval body, reviewers, compliance staff, and model owners can accept the workflow and its accountability controls."],"next_evidence_step":"Pre-register and run the authorized crossover simulation using matched synthetic or de-identified historical cases: compare the conventional packet with the proposed negative-space version on unresolved-hazard detection, correct classification of absent evidence, articulated dissent, retrieval success and delay, decision time, distress, and approval confidence. Falsify advancement if there is no meaningful improvement in the first three outcomes or if missed context, retrieval failure, distress, accessibility failure, or overconfident approval increases; no result may control a production deployment.","research_questions":["How often does the conventional review process produce the specified missed-hazard, ambiguous-absence, or suppressed-dissent failures?","Which component or combination—reversible isolation, explicit absence labels, protected interpretation, or chair-held silence—accounts for any observed change?","Can all omitted evidence remain immediately retrievable, complete, auditable, and accessible without undermining expert side-by-side comparison?","How do hierarchy, facilitation, disability access, and reviewer experience affect whether protected silence enables dissent or creates pressure?","What existing governance review formats overlap with this mechanism composition, and what incremental difference remains after that comparison?","What outcome thresholds would justify a reversible operational launch rather than abandonment or redesign?"],"recommendation":"PARTNERED_RESEARCH","uncertainty_constraints":["The packet provides no external evidence of problem prevalence, stakeholder demand, realized impact, or generalizability.","Prior art is unsearched, so novelty, rarity, and market differentiation are unmeasured.","The proposed causal effects are hypotheses; experienced panels may already perform reliably, and subtraction may conceal important dependencies.","Cost bands are resource-equivalent planning ranges rather than observed prices and depend strongly on review volume, systems, jurisdictions, and organizational scope.","The simulation can establish comparative performance under bounded conditions but cannot by itself establish safe production-wide adoption."],"closed_book_prior_art_boundary":"No external sources or prior-art search are included. The assessment treats the stated risk-ranking dashboard only as the packet's nearest rival and makes no claim about world novelty, prevalence, market size, or whether equivalent review practices already exist."}