{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp04_retrieval_first_paired20_20260802","cell_id":"layer_decay_and_expiration_management__education_pedagogy","round_index":0,"assessments":[{"hypothesis_id":"H1","search_queries":["site:education.nsw.gov.au curriculum previous syllabus archived current superseded","curriculum repository superseded version current curriculum materials official","Australian Curriculum version 8.4 archived website current version 9 official","site:australiancurriculum.edu.au archived version 8.4 current version 9"],"sources":[{"source_id":"H1-S1","title":"Australian Curriculum","publisher":"Australian Government Department of Education","url":"https://www.education.gov.au/australian-curriculum","source_class":"GOVERNMENT_OR_REGULATOR","claims_supported":["Version 9.0 is identified as the newer curriculum, while Version 8.4 remains available during jurisdictional transition.","Curriculum versions can coexist legitimately, so age alone cannot determine whether a version should be suppressed."]},{"source_id":"H1-S2","title":"Curriculum","publisher":"Australian Curriculum, Assessment and Reporting Authority","url":"https://www.acara.edu.au/curriculum","source_class":"OFFICIAL_ORGANIZATION_DATA","claims_supported":["ACARA publishes Version 9.0 on a new website and explicitly links the retained Version 8.4 website.","The official transition architecture distinguishes the successor from the older version while preserving access to both."]},{"source_id":"H1-S3","title":"Version history","publisher":"Australian Curriculum, Assessment and Reporting Authority","url":"https://www.australiancurriculum.edu.au/help/f-10-curriculum-overview/version-history","source_class":"OFFICIAL_GUIDANCE","claims_supported":["Curriculum changes are recorded through explicit version numbers and dates.","Downloaded and printed curriculum documents display their applicable version number, preserving version identity outside the repository."]},{"source_id":"H1-S4","title":"Improving K-12 Teachers’ Acceptance of Open Educational Resources by Open Educational Practices: A Mixed Methods Inquiry","publisher":"Journal of Education and Learning / PubMed Central","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC8445019/","source_class":"PRIMARY_RESEARCH","claims_supported":["Participating teachers reported outdated, broken, or missing OER links and difficulty determining resource quality and applicability.","The study establishes that stale educational resources impose real search and validation work on teachers."]}],"closest_analogue":"ACARA’s Version 8.4-to-9.0 curriculum transition, which assigns explicit version identities, designates a successor, links users to the newer website, and preserves the older version during a justified transition period.","overlap":"Explicit version identity, current-versus-older status, successor linkage, preserved historical access, and recognition that an older version may remain valid for some jurisdictions all overlap with lifecycle labeling and successor routing.","remaining_difference":"The analogue manages known formal versions, but the hypothesis additionally proposes semantic-staleness detection and retrieval-time demotion for distributed curriculum resources. A testable difference is whether that treatment reduces teachers’ selection of superseded resources without increasing median retrieval time compared with version labels and ordinary search alone.","classification":"POSSIBLE_DISTINCTION","disposition":"ADVANCE","rationale":"Official curriculum systems already implement strong version labeling and successor presentation, so the broad lifecycle idea is not new. The bounded distinction is automated semantic detection plus search-ranking intervention, evaluated on teacher selection errors and retrieval time; the shallow evidence does not establish that this combination has already been tested."},{"hypothesis_id":"H2","search_queries":["student knowledge model time decay evidence forgetting adaptive learning sequencing paper","Bayesian knowledge tracing forgetting time decay learner model primary research","time-aware knowledge tracing forgetting curve adaptive learning student model","ALEKS mastery evidence decay old assessment official"],"sources":[{"source_id":"H2-S1","title":"LF-LKT: A Logistic Regression Knowledge Tracing Model Integrating Learning and Forgetting","publisher":"International Society of the Learning Sciences","url":"https://repository.isls.org/handle/1/8902","source_class":"PRIMARY_RESEARCH","claims_supported":["LF-LKT weights learner evidence using time interval, recency, learning opportunities, and response information.","Across three datasets, the model improved predictive performance over AFM and PFA baselines."]},{"source_id":"H2-S2","title":"Rethinking and Improving Student Learning and Forgetting Processes for Attention based Knowledge Tracing Models","publisher":"AAAI Conference on Artificial Intelligence","url":"https://ojs.aaai.org/index.php/AAAI/article/view/34998","source_class":"PRIMARY_RESEARCH","claims_supported":["Existing knowledge-tracing models already use time-decay matrices for historical interactions.","LefoKT adds relative-forgetting attention and reports validation across three datasets."]},{"source_id":"H2-S3","title":"ALEKS Instructor’s Manual","publisher":"ALEKS Corporation","url":"https://www.aleks.com/manual/pdf/educators-hssl.pdf","source_class":"COMMERCIAL_FIRST_PARTY","claims_supported":["ALEKS automatically reassesses learners based on progress and elapsed time, including a periodic 60-day trigger.","Recent learning history receives special emphasis, reassessment updates readiness, and topics lost on a recent assessment are routed back into instruction while prior assessment reports remain available."]},{"source_id":"H2-S4","title":"Capturing Session-to-Session Dynamics of Learning and Forgetting: Testing the Limits of Knowledge Tracing Models","publisher":"International Journal of Artificial Intelligence in Education","url":"https://www.sciencedirect.com/science/article/pii/S1560429226000776","source_class":"PRIMARY_RESEARCH","claims_supported":["The study compares knowledge-tracing models with and without forgetting using longitudinal data.","It finds that acceptable retrospective fit does not guarantee reliable future prediction and identifies failures involving spacing and forgetting."]}],"closest_analogue":"LF-LKT and related forgetting-aware knowledge-tracing models, supplemented by ALEKS’s deployed cycle of time/progress-triggered reassessment and readiness updates.","overlap":"Historical learner observations receive time-sensitive or recency-sensitive influence; learner state is continually updated; recent validated performance affects subsequent activity readiness; and historical reports can be retained separately from current instructional state.","remaining_difference":"The exact decision-level test—whether decaying or expiring old negative observations reduces obsolete remediation without increasing premature advancement—remains useful validation, as does separating hard instructional expiry from longitudinal retention. These are evaluation and policy refinements of an established causal lever.","classification":"OBVIOUS_COLLISION","disposition":"REJECT","rationale":"Multiple primary models already decay historical learner evidence, and a deployed adaptive platform already uses elapsed time, recent evidence, reassessment, and changed readiness to revise sequencing. The proposed outcome comparison is worthwhile research but does not preserve enough mechanistic separation for this screen."},{"hypothesis_id":"H3","search_queries":["course Q&A archive answers previous semester reuse Piazza official","Ed Discussion duplicate course previous semester posts official","research reuse discussion forum posts previous course offering question answering students","course question answer knowledge base across semesters stale answers study"],"sources":[{"source_id":"H3-S1","title":"Instructors: Clone a Class","publisher":"Piazza Technologies","url":"https://support.piazza.com/support/solutions/articles/48000616706-instructors-clone-a-class","source_class":"OFFICIAL_PRODUCT_DOCUMENTATION","claims_supported":["Piazza creates a new class for a new term while allowing course information, settings, folders, and resources to be carried over.","Prior posts are not activated wholesale; instructors copy individual posts into the new class."]},{"source_id":"H3-S2","title":"Guide to Course Reruns","publisher":"Open edX Documentation Working Group / Axim Collaborative","url":"https://docs.openedx.org/en/latest/educators/references/course_development/course_reruns.html","source_class":"OFFICIAL_PRODUCT_DOCUMENTATION","claims_supported":["A course rerun creates an independent course instance and requires dates to be updated.","Discussion posts, responses, and comments are not copied into the rerun, creating a default course-context boundary for old answers."]},{"source_id":"H3-S3","title":"In-video reuse of discussion threads in MOOCs","publisher":"University of Minnesota Research Repository","url":"https://experts.umn.edu/en/publications/in-video-reuse-of-discussion-threads-in-moocs/","source_class":"PRIMARY_RESEARCH","claims_supported":["The study evaluated reuse of selected high-value discussion threads from past MOOC sessions in future sessions.","Survey evidence found the reused threads useful to many learners and examined the role of instructor involvement."]},{"source_id":"H3-S4","title":"Managing an Online Forum","publisher":"MIT OpenCourseWare","url":"https://ocw.mit.edu/courses/6-034-artificial-intelligence-fall-2010/pages/instructor-insights/managing-an-online-forum/","source_class":"OFFICIAL_GUIDANCE","claims_supported":["Course staff describe prior-year forum material as an archive used to identify recurring learner difficulties.","The archived material supports staff work, but the page does not describe automatic cross-year answer retrieval for current students."]}],"closest_analogue":"Piazza’s term-specific class cloning with individual post carryover: old Q&A remains associated with the former class, and an instructor must selectively copy a post before it becomes active in the new class.","overlap":"Course-run context acts as a lifetime boundary; old discussions remain preserved; bulk reactivation is avoided; and selective copying functions as a manual renewal gate. MOOC research also demonstrates deliberate reuse of selected past threads.","remaining_difference":"The hypothesis applies expiration and renewal at query time within a searchable archive rather than through whole-course boundaries and manual copying. The bounded test is whether automatic context checks reduce obsolete-answer impressions while preserving resolution of unchanged questions and avoiding duplicate-question growth.","classification":"POSSIBLE_DISTINCTION","disposition":"ADVANCE","rationale":"Existing systems already provide a close manual analogue, so mere course scoping or selective reuse is not distinctive. The potentially separable treatment is answer-level, query-time context expiration with measurable misinformation and resolution outcomes; the opened evidence does not show that treatment evaluated directly."},{"hypothesis_id":"H4","search_queries":["official assessment item bank retire item preserve historical attempts dependencies","Moodle question bank delete question used quiz attempts official documentation","Questionmark retire question status item bank official documentation","Canvas item bank delete question used assessment preserve results official"],"sources":[{"source_id":"H4-S1","title":"Question bank","publisher":"MoodleDocs","url":"https://docs.moodle.org/405/en/Question_bank","source_class":"OFFICIAL_PRODUCT_DOCUMENTATION","claims_supported":["Moodle exposes question versions, usage counts, last-used information, and readiness or review status.","If a question marked for deletion is still used by a quiz, Moodle retains it in the database and hides it rather than destructively deleting it."]},{"source_id":"H4-S2","title":"Managing questions","publisher":"MoodleDocs","url":"https://docs.moodle.org/405/en/Managing_questions","source_class":"OFFICIAL_PRODUCT_DOCUMENTATION","claims_supported":["Moodle reports where question types are used and records dependencies between question types.","Making a type unavailable prevents new creation while preserving existing quizzes and review of old attempts; uninstalling is allowed only when usage and dependency conditions are clear."]},{"source_id":"H4-S3","title":"Deleting Questions and Reappearing","publisher":"Instructure Community","url":"https://community.instructure.com/en/discussion/609590/deleting-questions-and-reappearing/p1","source_class":"COMMERCIAL_FIRST_PARTY","claims_supported":["Mastery Connect retains an item when it has already been used in an assessment and is attached to a tracker containing student data.","The stated purpose is preservation of data integrity rather than allowing destructive removal."]}],"closest_analogue":"Moodle’s question-bank logical deletion: usage is inspectable, and a question still referenced by a quiz is hidden from ordinary selection but retained so attempts and downstream behavior do not break.","overlap":"Inbound-use checks, version and usage metadata, exclusion from future use, logical rather than destructive deletion, preservation of completed attempts, and refusal to remove depended-upon components closely match dependency-gated quarantine.","remaining_difference":"The hypothesis broadens the dependency graph to forms, reports, scoring lineage, and appeals, and proposes measuring decision-to-exclusion time and broken references. It may also add an explicit time-bounded quarantine and automated dependency-resolution workflow.","classification":"OBVIOUS_COLLISION","disposition":"REJECT","rationale":"Current assessment platforms already implement the central mechanism: dependency-sensitive refusal of destructive deletion combined with hidden or unavailable status that preserves historical evidence. Wider dependency coverage and comparative timing metrics are implementation and evaluation extensions rather than a distinct treatment."},{"hypothesis_id":"H5","search_queries":["university student records digital archive restore testing policy official","education records retention archive disaster recovery test restoration official","higher education digital records storage tiering archive retrieval accreditation official","student record archive backup restoration testing registrar"],"sources":[{"source_id":"H5-S1","title":"Record Management Policy","publisher":"Leaders Institute","url":"https://leaders.edu.au/policies_/record-management-policy/","source_class":"OFFICIAL_ORGANIZATION_DATA","claims_supported":["Electronic records are appraised before transition to inactive storage, permanent retention, or disposal.","Inactive records are archived, external storage requires a catalogue, student records are backed up, and the policy requires records to remain accessible and managed through their lifecycle.","The policy also specifies regular retrieval testing for research records."]},{"source_id":"H5-S2","title":"Electronic Backup Policy","publisher":"Spalding University","url":"https://policyguide.spalding.edu/?id=4698","source_class":"OFFICIAL_ORGANIZATION_DATA","claims_supported":["The university uses different backup media and retention periods and publishes restoration-time targets.","Backup testing is performed by restoring files regularly, with restored-file logging used to verify backup integrity."]},{"source_id":"H5-S3","title":"Vital Records and Records Disaster Mitigation and Recovery","publisher":"U.S. National Archives and Records Administration","url":"https://www.archives.gov/records-mgmt/vital-records","source_class":"GOVERNMENT_OR_REGULATOR","claims_supported":["NARA calls for explicit records-recovery procedures and periodic tests of recovery plans.","Testing should evaluate recovery activities and drive modifications to plans and responsibilities."]},{"source_id":"H5-S4","title":"DfE retention schedule","publisher":"UK Department for Education","url":"https://www.gov.uk/government/publications/retention-schedule-for-data-processed-by-dfe/dfe-retention-schedule","source_class":"GOVERNMENT_OR_REGULATOR","claims_supported":["The schedule assigns differentiated retention, review, archival, access-restriction, and destruction paths to learner and education records.","Some learning-achievement records move to archive after active use and are periodically reviewed, demonstrating lifecycle transitions for education data."]}],"closest_analogue":"Leaders Institute’s records-management lifecycle, which transitions appraised electronic records to inactive storage or archive while retaining cataloguing, accessibility, backup, and retrieval-testing obligations.","overlap":"Education-record classification, active-to-inactive archival transition, differentiated retention, external-storage catalogues, preserved accessibility, backup, restore procedures, periodic recovery testing, and explicit recovery targets collectively cover nearly the full proposed lifecycle.","remaining_difference":"The hypothesis specifies automated age-and-access hot/warm/cold placement for completed-course bundles and stratified restore sampling across formats and age bands, with predefined restoration-rate and latency metrics. Those are more precise operational controls than the cited policies describe.","classification":"OBVIOUS_COLLISION","disposition":"REJECT","rationale":"Educational institutions and records authorities already combine inactive-record archiving, differentiated retention, restoration procedures, and recurring restore tests. Automated tier selection and stratified sampling improve execution but do not create a sufficiently distinct intervention for advancement in this shallow screen."}],"nominated_ids":["H1","H3"],"replenishment_recommended":false}