{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp09_archetype_breadth150_20260804","research_id":"eoa_inverse_innovation_exp09_light_prior_art_20260804","cell_id":"modular_decomposition__computer_science","search_lanes":{"direct_problem_and_intervention":{"queries":["machine learning data deletion workflow identity resolution data lineage model artifacts audit evidence","machine unlearning data deletion pipeline data lineage audit compliance","GDPR deletion request ML models lineage deletion workflow architecture","data deletion pipeline subject identity matching lineage remediation audit microservices"],"source_ids":["SRC1","SRC2","SRC3"],"no_result_note":"No retained source directly documented the proposed monolithic-worker failure or the exact four-module intervention."},"synonyms_and_historical_terms":{"queries":["machine unlearning pipeline architecture data lineage removal request audit trail paper","machine forgetting algorithmic forgetting selective forgetting data removal","data subject request deletion workflow identity verification data discovery audit trail","right to erasure architecture data lineage deletion orchestration"],"source_ids":["SRC1","SRC2","SRC3"],"no_result_note":null},"products_practices_and_standards":{"queries":["privacy request orchestration identity verification data mapping deletion evidence product","Data Subject Request automation identity verification discovery deletion audit evidence","data rights automation identity correlation deletion validation auditable evidence","AWS decompose monolith business capability subject matter experts"],"source_ids":["SRC1","SRC2","SRC4"],"no_result_note":null},"component_combination":{"queries":["software architecture data deletion workflow microservices contract manifest audit completion invariant","microservices versioned contracts ownership integration invariants audit workflow modular decomposition","machine unlearning data deletion verification audit pipeline","decompose monolith responsibility boundaries independently testable modules"],"source_ids":["SRC1","SRC3","SRC4"],"no_result_note":"The component combination was visible across adjacent sources, but no retained source specified Subject Resolution, Artifact-Lineage Closure, Remediation Planning, and Completion Evidence as four stewarded modules joined by versioned manifests and a coverage invariant."}},"sources":[{"source_id":"SRC1","title":"Data Rights Automation Software: DSAR & Deletion Requests","publisher":"BigID","url":"https://bigid.com/data-rights-automation-app/","source_type":"FIRST_PARTY_PRODUCT","claims_supported":["Commercial deletion-request automation already combines identity correlation, personal-data discovery, task routing, deletion execution or integration, validation, remediation tracking, auditable evidence, and documented completion.","The vendor identifies identity gaps, disconnected privacy tools, deletion uncertainty, manual routing, and incomplete reporting as operational problems."]},{"source_id":"SRC2","title":"Data Subject Request (DSR) Automation","publisher":"OneTrust","url":"https://www.onetrust.com/products/data-subject-request-dsr-automation/","source_type":"FIRST_PARTY_PRODUCT","claims_supported":["A commercial DSR workflow already automates intake, identity verification, personal-data discovery, deletion, legal-hold checks, redaction, and secure response.","The vendor characterizes DSR fulfillment as complex, time-consuming, and composed of numerous manual tasks."]},{"source_id":"SRC3","title":"Forget Unlearning: Towards True Data-Deletion in Machine Learning","publisher":"Proceedings of Machine Learning Research","url":"https://proceedings.mlr.press/v202/chourasia23a.html","source_type":"PRIMARY_RESEARCH","claims_supported":["Deleting a person's stored records does not by itself remove their influence from trained models.","ML deletion guarantees can fail because records and model computations are interdependent, and cached partial computations can leak deleted information across releases."]},{"source_id":"SRC4","title":"Decompose by business capability","publisher":"Amazon Web Services","url":"https://docs.aws.amazon.com/prescriptive-guidance/latest/modernization-decomposing-monoliths/decompose-business-capability.html","source_type":"OFFICIAL_GUIDANCE","claims_supported":["Decomposing a monolith around business capabilities with relevant subject-matter experts is an established architecture pattern.","Claimed advantages include loose coupling and teams organized around business value, while identifying correct capabilities requires substantial domain understanding."]}],"problem_evidence":{"status":"PARTLY_SUPPORTED","finding":"Public evidence makes the broader problem visible: deletion workflows span identity verification, discovery, fulfillment, validation, and evidence; vendors report identity gaps, siloed tools, manual coordination, deletion uncertainty, and incomplete reporting; and primary research shows that deletion becomes harder when ML models and cached computations are included. The specific empirical assertion that these responsibilities reside in one shared-state worker whose local changes cause broad reviews and cross-stage defects was not independently demonstrated.","source_ids":["SRC1","SRC2","SRC3"]},"closest_prior_art":[{"name":"BigID Data Rights Automation","source_ids":["SRC1"],"overlap":"Covers nearly the same functional chain: identity correlation, data discovery and mapping, task routing, deletion integration, validation, remediation tracking, auditable evidence, and documented completion.","remaining_difference":"The public page does not disclose the candidate's four internal responsibility boundaries, encapsulated representations, named engineering stewards, versioned interstage manifests, independent module tests, lineage-snapshot compatibility checks, or the exact artifact-disposition completion invariant."},{"name":"OneTrust DSR Automation","source_ids":["SRC2"],"overlap":"Automates a staged deletion-request workflow incorporating identity verification, discovery, deletion, legal checks, redaction, and response.","remaining_difference":"It does not publicly specify ML artifact-lineage closure, model-remediation planning, four responsibility-coherent code modules, contract-version checks, or the proposed cross-module completion invariant."},{"name":"Business-capability monolith decomposition applied to an ML deletion workflow","source_ids":["SRC1","SRC3","SRC4"],"overlap":"The combination supplies an established decomposition pattern, a commercially established end-to-end deletion workflow, and the ML-specific requirement to address model influence rather than database records alone.","remaining_difference":"No retained source directly combines those elements into the exact four-module, stewarded, manifest-based architecture or tests whether that design reduces local-change review surface while preserving request-level agreement."}],"prior_art_disposition":"ADJACENT_PRIOR_ART","contrastive_claim_remaining":"Compared with established DSR orchestration and generic business-capability decomposition, the remaining falsifiable claim is that placing subject resolution, ML artifact-lineage closure, remediation planning, and completion evidence behind four stewarded internal boundaries with versioned manifests and an artifact-coverage completion invariant reduces the code, owners, and tests touched by representative local changes while producing exactly the same downstream manifests and completion decisions as the baseline. The claim is about internal change tractability plus cross-stage coherence, not the already-established idea of staged deletion automation.","contrastive_claim_falsifier":"Falsify the claim if dependency tracing shows that the four responsibilities must routinely co-design or atomically mutate one representation, if the current worker already provides stable independently testable handoffs, or if shadow extraction of Subject Resolution requires unrelated remediation or evidence changes and adds contract or coordination work without reducing review surface or integration mismatches.","gates":{"adequate_source_search":{"status":"PASS","rationale":"The bounded search covered direct phrasing, machine-unlearning and right-to-erasure terminology, commercial DSR products, monolith-decomposition guidance, and combinations of identity, lineage, remediation, verification, and audit components. Four opened sources span four publishers and include two first-party products, primary research, and official guidance.","source_ids":["SRC1","SRC2","SRC3","SRC4"]},"supported_problem":{"status":"PASS","rationale":"The broader multi-responsibility deletion and ML-remediation problem is supported, while the proposal-specific monolithic implementation and blast-radius assertions remain unverified; therefore the evidence is appropriately PARTLY_SUPPORTED.","source_ids":["SRC1","SRC2","SRC3"]},"distinct_testable_claim":{"status":"PASS","rationale":"Although the functional workflow and decomposition principle have adjacent prior art, the narrower claim about four internal boundaries, versioned manifests, independent change surface, and preserved completion coherence is distinct and measurable.","source_ids":["SRC1","SRC2","SRC4"]},"bounded_next_test":{"status":"PASS","rationale":"The proposed test is limited to at most 20 sanitized or synthetic traces, five change scenarios, and a disposable Subject Resolution adapter. It measures undeclared dependencies, independent testability, review surface, and exact downstream equivalence without estimating production effect size.","source_ids":["SRC1","SRC3","SRC4"]},"no_obvious_safety_or_authority_stop":{"status":"PASS","rationale":"A non-production disposable branch and sanitized or synthetic shadow replay avoid production deletion, retraining, status changes, and direct-identifier exposure. Production policy and completion authority remain with privacy, data, and compliance owners, with explicit halt and rollback conditions.","source_ids":["SRC1","SRC2","SRC3"]}},"screen_survival":true,"world_novelty_boundary":"This coarse, bounded public-web screen found adjacent functional and architectural prior art but no direct disclosure of the exact combined implementation. It cannot establish world novelty, patentability, freedom to operate, market size, expert acceptance, production feasibility, effect size, or realized value."}