{"schema_version":1,"research_id":"eoa_inverse_innovation_exp05_external_evaluation_20260803","source_assessment_id":"layer_decay_and_expiration_management__cognitive_science:P5:v0","cell_id":"layer_decay_and_expiration_management__cognitive_science","search_queries":["site:ies.ed.gov practice guide scaffolding fade prompts worked examples independent practice procedural learning","instructional scaffolding fading prompts cognitive tutor primary research adaptive fading assistance dilemma","worked examples fading guidance effect primary study cognitive load","accessibility accommodation removing scaffolds legal ADA online training official guidance","Koedinger Aleven Exploring assistance dilemma experiments intelligent tutoring systems 2007 PDF","site:adlnet.gov xAPI specification state activity profile context official","site:ed.gov disability accommodations auxiliary aids education official online instruction","intelligent tutoring system scaffold fading learner control adaptive scaffolding primary study assistance fading","ADL xAPI specification official State Resource Activity Profile API","IEEE learning technology standards learning object metadata educational support history configuration","IMS Caliper Analytics official specification event context learning activity","NIST records retention deletion audit archival restore official","\"Does working memory moderate the effect of fading on math performance\" full text","\"Does Knowing When Help Is Needed Improve Subgoal Hint Performance\" PDF","site:pmc.ncbi.nlm.nih.gov PMC12879535","10.1609/aaai.v37i13.26887 PDF"],"sources":[{"source_id":"S1","title":"Cognitive-Load Theory: Methods to Manage Working Memory Load in the Learning of Complex Tasks","publisher":"SAGE Publications, Current Directions in Psychological Science","url":"https://journals.sagepub.com/doi/10.1177/0963721420922183","source_class":"AUTHORITATIVE_SECONDARY","publication_date":"2020","accessed_at":"2026-08-03","claims_supported":["Multiple distributed information sources increase attentional switching and extraneous load.","Worked examples can benefit novices, while becoming redundant and imposing unnecessary load for more advanced learners.","Guidance should change as domain expertise develops."]},{"source_id":"S2","title":"Exploring the Assistance Dilemma in Experiments with Cognitive Tutors","publisher":"Springer / Carnegie Mellon University","url":"https://pact.cs.cmu.edu/koedinger/pubs/paper%20%2314.pdf","source_class":"PRIMARY_RESEARCH","publication_date":"2007-07-21","accessed_at":"2026-08-03","claims_supported":["Cognitive Tutors already provide step-level feedback, error-specific messages, on-demand hints, adaptive problem selection, and detailed event logging.","Balancing assistance giving and withholding is an established instructional-science problem.","The authors conclude that the conditions and parameters governing assistance require a broader empirical base."]},{"source_id":"S3","title":"Static, Dynamic, or Adaptive? Comparing Interface Scaffolding Methods in an Intelligent Tutoring System","publisher":"Georgia Institute of Technology, Teachable AI Lab","url":"https://tail.cc.gatech.edu/publications/reddig-pele-2026/","source_class":"PRIMARY_RESEARCH","publication_date":"2026","accessed_at":"2026-08-03","claims_supported":["A technical college deployed full, mastery-adaptive, and on-demand scaffolding across nine College Algebra sections.","Dynamic conditions outperformed full scaffolding on reported course outcomes; the on-demand condition produced significantly higher grades and substantially more completed problems.","Full-to-adaptive and full-to-on-demand interface transitions are close operational prior art."]},{"source_id":"S4","title":"Organizing Instruction and Study to Improve Student Learning","publisher":"U.S. Department of Education, Institute of Education Sciences, What Works Clearinghouse","url":"https://ies.ed.gov/ncee/wwc/practiceguide/1","source_class":"GOVERNMENT_OR_REGULATOR","publication_date":"2007-09","accessed_at":"2026-08-03","claims_supported":["The official guide recommends alternating worked solutions with independent problem solving, with moderate evidence.","The guide recommends active retrieval and tests to identify material needing further study.","Its intended audiences include administrators, policymakers, researchers, specialists, and teachers, identifying institutional authorizers and adopters for instructional changes."]},{"source_id":"S5","title":"Disability Discrimination: Academic Adjustments for Postsecondary Students","publisher":"U.S. Department of Education, Office for Civil Rights","url":"https://www.ed.gov/laws-and-policy/civil-rights-laws/disability-discrimination/disability-discrimination-key-issues/disability-discrimination-academic-adjustments-postsecondary-students","source_class":"OFFICIAL_GUIDANCE","publication_date":"2025-01-14","accessed_at":"2026-08-03","claims_supported":["Eligible postsecondary students may require academic modifications and auxiliary aids or services to ensure equal opportunity.","An ordinary instructor or automated tutor cannot safely treat authorized disability-related support as a disposable pedagogical scaffold."]},{"source_id":"S6","title":"IEEE 1484.12.1-2020: IEEE Standard for Learning Object Metadata","publisher":"IEEE Standards Association","url":"https://standards.ieee.org/ieee/1484.12.1/7699/","source_class":"STANDARD","publication_date":"2020-11-16","accessed_at":"2026-08-03","claims_supported":["An active standard defines metadata for digital or non-digital learning objects.","The schema includes lifecycle, educational, technical, rights, relation, annotation, and classification categories.","Learning-object metadata can support management, location, evaluation, exchange, and adaptive scheduling, making a scaffold inventory technically plausible but not proving dependency checks or restoration."]},{"source_id":"S7","title":"IRT-Based Adaptive Hints to Scaffold Learning in Programming","publisher":"ERIC / IEEE Transactions on Learning Technologies","url":"https://eric.ed.gov/?id=EJ1200741","source_class":"AUTHORITATIVE_SECONDARY","publication_date":"2018","accessed_at":"2026-08-03","claims_supported":["A probabilistic intelligent-tutoring system varied assistance based on predicted success.","The reported comparison found best performance at a specified success-probability target and automatically reduced hints as learners grew.","Performance-triggered adaptive hint fading is substantial prior art."]},{"source_id":"S8","title":"Effects of Faded Scaffolding in Computer-Based Instruction on Learners' Performance, Cognitive Load, and Test Anxiety","publisher":"ERIC / ProQuest LLC","url":"https://eric.ed.gov/?id=ED570610","source_class":"AUTHORITATIVE_SECONDARY","publication_date":"2016","accessed_at":"2026-08-03","claims_supported":["A study compared gradual and sudden scaffold fading in computer-based instruction.","Among the analyzed participants, gradual fading did not significantly outperform sudden fading on performance, cognitive load, or test anxiety.","Fixed fading may be poorly individualized, and gradual withdrawal cannot be presumed superior to a simpler comparator."]}],"problem_evidence":{"support":"MODERATE","rationale":"The general problem visibly exists: guidance that helps novices can become redundant and impose load as expertise develops, and balancing help with independent work is an established dilemma. A 2026 technical-college deployment also found better outcomes under adaptive or on-demand interfaces than under full scaffolding. However, none of the eight sources measures the proposal's specific prevalence claim—multiple independently deposited, stale scaffolds simultaneously accumulating within procedural-training modules—so local magnitude remains unknown.","source_ids":["S1","S2","S3","S8"]},"stakeholder_evidence":{"support":"MODERATE","rationale":"Identifiable adopters and authorizers exist. A technical college deployed alternative scaffold states across nine sections, while the IES guide addresses teachers, administrators, policymakers, and researchers and recommends combining worked examples with independent problem solving. The Department of Education identifies an accessibility authority constraint. These sources express need for calibrated guidance and equal access, but not demand for the complete lifecycle-governance bundle.","source_ids":["S3","S4","S5"]},"prior_art":{"proximity":"SUBSTANTIAL_COLLISION","closest_analogues":[{"name":"Full, mastery-adaptive, and on-demand Apprentice Tutors interfaces","similarity":"Directly compares persistent fine-grained steps with mastery-triggered fading and learner-requestable support in a live postsecondary deployment.","remaining_difference":"It does not report per-scaffold review leases, reference dependency gates, accommodation exceptions, quarantine rollback, tombstones, or restoration of historical support configurations.","source_ids":["S3"]},{"name":"IRT-based adaptive hint fading","similarity":"Automatically reduces instructional hints based on a learner-performance model and empirically compares assistance policies.","remaining_difference":"It optimizes current hint quantity rather than governing the lifecycle, ownership, staleness, institutional dependencies, and archival history of each scaffold.","source_ids":["S7"]},{"name":"Cognitive Tutor assistance management","similarity":"Established systems provide step-specific feedback, error messages, on-demand hints, adaptive selection, and detailed logging while explicitly addressing assistance giving versus withholding.","remaining_difference":"The literature frames an assistance-policy problem, not an auditable retirement workflow with dependency checks and reversible content states.","source_ids":["S2"]},{"name":"IEEE Learning Object Metadata lifecycle representation","similarity":"Provides standardized lifecycle, educational, technical, rights, relation, and annotation metadata for learning objects.","remaining_difference":"It supplies representation primitives but no scaffold-withdrawal decision rule, live dependency proof, accommodation authority workflow, probe, rollback policy, or archive-restore test.","source_ids":["S6"]},{"name":"Gradual versus sudden scaffold fading","similarity":"Directly tests staged and abrupt withdrawal of computer-based scaffolds.","remaining_difference":"It lacks individualized dependency gating and lifecycle records; it also found no significant advantage for gradual fading, making it a serious falsifying comparator.","source_ids":["S8"]}],"distinctive_claim_remaining":"For individually identified scaffolds in one procedural-training module, an instructor- and accessibility-authorized lifecycle workflow—semantic-staleness review, explicit instructional and accommodation dependency checks, matched scaffold-reduced probes, visible-to-requestable-to-hidden transitions, time-bounded rollback, and reconstructable configuration records—will improve the validity and interpretability of claims about independent procedural competence relative to both the existing configuration and a fixed-stage or mastery-only fading policy, without increasing critical errors, unauthorized support loss, or review burden beyond prespecified limits.","confidence":"HIGH"},"implementation_evidence":{"support":"MODERATE","rationale":"Existing tutors already implement adaptive hints, on-demand access, performance-triggered fading, interface experiments, and detailed event logging. IEEE metadata supplies lifecycle and relation fields suitable for a registry. A read-only inventory and sandboxed state transitions are therefore technically credible. Evidence is absent for complete automatic dependency discovery, reliable reconstruction after interface changes, privacy protection for learner-linked configuration bundles, human review effort, integration with a particular LMS, and safe treatment of accommodations.","source_ids":["S2","S3","S5","S6","S7","S8"]},"scores":{"meaningful_impact":{"score":3,"rationale":"Misattributing cue-supported completion to independent competence can matter, and redundant guidance can impose load, but prevalence and effect size for accumulated scaffolds are unmeasured.","source_ids":["S1","S2","S3"]},"stakeholder_pull":{"score":2,"rationale":"Institutions demonstrably deploy and compare scaffold policies, and IES addresses responsible adopters, but no source requests the proposed lifecycle-governance system.","source_ids":["S3","S4"]},"incremental_advantage":{"score":2,"rationale":"The integrated governance and audit layer may add interpretability, but no comparison establishes an advantage over adaptive, on-demand, fixed-stage, or instructor-controlled fading; gradual fading has mixed evidence.","source_ids":["S3","S7","S8"]},"distinctiveness_plausibility":{"score":3,"rationale":"The exact combination of dependency gating, accommodation authority, reversible tiers, tombstones, and tested historical reconstruction was not found, although its central fading and metadata components substantially collide with established work.","source_ids":["S2","S3","S6","S7","S8"]},"technical_implementability":{"score":4,"rationale":"Metadata registries, adaptive support states, logging, and on-demand interfaces are established technical capabilities. Semantic staleness inference and exhaustive dependency checks remain difficult.","source_ids":["S2","S3","S6","S7"]},"adoption_authority_feasibility":{"score":3,"rationale":"Instructors, designers, administrators, and accessibility officials are identifiable authorities, but shared approval and accommodation handling add governance friction.","source_ids":["S4","S5"]},"evidence_readiness":{"score":3,"rationale":"A bounded three-condition sandbox study can be run using existing tutoring methods, but it requires a live module, learner participation, institutional data, and authorized personnel.","source_ids":["S2","S3","S7","S8"]},"safety_net_benefit":{"score":3,"rationale":"Requestable and recoverable states can reduce harm from mistaken withdrawal, while explicit accommodation control is essential. Retaining configuration bundles introduces privacy and unauthorized-access risks not yet tested.","source_ids":["S5","S6"]},"scalability":{"score":3,"rationale":"Standard metadata and automated state transitions could scale, but reference tracing, semantic review, accommodation adjudication, and restore testing may remain labor-intensive.","source_ids":["S2","S6"]}},"score_confidence":"MODERATE","costs":{"first_evidence":{"band_2026_usd":"10K_TO_50K","scope":"Six-to-ten-week shadow audit and formative sandbox study for one non-safety-critical module, approximately 10–30 scaffolds, matched task forms, three comparators, accessibility review, basic analysis, and one historical-render test.","confidence":"LOW","assumptions":["Existing tutor and sandbox are available.","No production curriculum, summative scores, accommodations, or source records are changed.","Resource-equivalent estimate assumes part-time effort from an instructor, instructional designer, engineer or analyst, accessibility reviewer, and research coordinator.","Participant recruitment, consent or IRB determination, and compensation remain small-scale."],"source_ids":["S2","S3","S5","S6"]},"initial_deployment_startup":{"band_2026_usd":"50K_TO_250K","scope":"Production-quality registry and lifecycle workflow for one platform and one to three modules, including roles, audit logging, support-state transitions, migration, privacy/accessibility review, and rollback tooling.","confidence":"LOW","assumptions":["The LMS or tutor exposes content identifiers, configuration APIs, and event logs.","Dependency mapping begins with declared references and human review rather than complete automated discovery.","Existing identity, authorization, backup, and analytics services are reused."],"source_ids":["S2","S5","S6"]},"operational_launch":{"band_2026_usd":"250K_TO_1M","scope":"Institutional launch across approximately 5–15 procedural modules, including content inventory, integrations, training, governance approvals, monitoring, restore drills, evaluation, and change management.","confidence":"LOW","assumptions":["Launch occurs at one institution on one principal learning platform.","Safety-critical training and high-stakes summative assessment remain excluded initially.","Substantial instructional-design and engineering labor is required to classify legacy scaffolds and dependencies."],"source_ids":["S3","S5","S6"]},"annual_recurring":{"band_2026_usd":"50K_TO_250K","scope":"Annual review operations for an institutional deployment: policy updates, stale-scaffold triage, accommodation and privacy review, platform maintenance, audit-log retention, restore sampling, and outcome monitoring.","confidence":"LOW","assumptions":["Approximately 0.5–1.5 combined FTE plus hosting and periodic specialist review.","Most state transitions are automated after authorization.","Major platform replacement or broad content redevelopment is excluded."],"source_ids":["S5","S6"]}},"verified_pipeline_gates":{"externally_supported_problem":{"status":"YES","reason":"Independent research and a live deployment support the assistance dilemma, expertise-dependent redundancy, and possible disadvantages of leaving all scaffolding visible, although local accumulation prevalence is not established.","source_ids":["S1","S2","S3"]},"externally_credible_adopter_or_authorizer":{"status":"YES","reason":"A technical college is an observed adopter of scaffold-policy experimentation; instructors and administrators are named IES audiences; the Department of Education establishes the relevance of accessibility authority.","source_ids":["S3","S4","S5"]},"distinct_testable_incremental_claim":{"status":"YES","reason":"The residual claim contrasts dependency-gated, reversible, reconstructable lifecycle governance against current support, fixed-stage fading, and mastery-only or on-demand rivals using observable performance, restoration, error, authority, and workload outcomes.","source_ids":["S3","S6","S7","S8"]},"bounded_next_evidence_step":{"status":"YES","reason":"One module and a reviewed scaffold subset can be tested in a formative sandbox with matched task forms, three conditions, predefined stop rules, and a reconstruction drill.","source_ids":["S2","S3","S8"]},"no_unresolved_safety_or_authority_stop":{"status":"YES","reason":"The proposed first step is non-safety-critical, read-only outside the sandbox, and preserves production settings and records. Proceeding requires instructor approval and accessibility-authority control over accommodation exceptions; automated withdrawal is not authorized.","source_ids":["S5"]},"credible_cost_scope_and_range":{"status":"UNCERTAIN","reason":"Scopes and resource-equivalent ranges are bounded, but no selected source reports costs for building or operating this exact lifecycle workflow; integration quality and manual review effort could move deployment by more than one band.","source_ids":[]}},"next_evidence_step":"Partner with one instructor, instructional designer, platform operator, and accessibility authority for one existing non-safety-critical procedural module. First inventory 10–30 visible or requestable scaffolds and record owner, purpose, covered steps, age, declared relations, current status, withdrawal rule, and accommodation status. Falsify the problem if the audit finds no superseded, duplicated, unowned, reference-broken, or indefinitely active scaffold and matched supported versus reduced probes reveal no hidden dependence. Then preregister a small formative sandbox comparison using matched task forms: (A) current support configuration, (B) fixed-stage or existing mastery-adaptive fading, and (C) dependency-gated visible-to-requestable-to-hidden lifecycle for a reviewed subset. Measure unsupported task-step accuracy, latency, critical omissions, help requests, strategy explanations, cognitive-load rating, stop-condition activations, restoration events, reviewer disagreement, per-scaffold review time, and successful rendering of one archived prior configuration. Halt and restore the prior configuration upon any accommodation loss, safety-content classification, broken live reference, critical-error threshold crossing, or reconstruction failure. Falsify the incremental claim if condition C does not improve prespecified independent-performance attribution or outcome measures relative to the best rival, causes more critical errors or unauthorized support loss, repeatedly restores supposedly retireable aids, fails reconstruction, or exceeds the prespecified review-effort ceiling. Do not alter production settings, summative scores, accommodations, or source records.","blocking_evidence":["No externally measured prevalence of stale, duplicated, unowned, or indefinitely active scaffolds within the target kind of procedural module.","No comparative evidence that the complete lifecycle bundle outperforms adaptive, on-demand, fixed-stage, or instructor-controlled fading.","No validated measure of whether successful performance is independently generated rather than scaffold-dependent in the proposed workflow.","No field evidence that declared and observed dependency tracing is sufficiently complete.","No demonstrated accommodation-review process for the intended institution and learner population.","No privacy, access-control, retention, or reconstruction assessment for learner-linked historical support bundles.","No measured instructor, designer, accessibility-reviewer, or engineering workload; cost bands remain assumption-driven.","No live test of rollback thresholds, learner willingness to request warm-tier help, or interface consistency after repeated state transitions."],"research_disposition":"PARTNERED_RESEARCH_PROGRAM","world_novelty_boundary":"The search establishes substantial prior art for guidance fading, adaptive and on-demand hints, assistance-policy experiments, event logging, and lifecycle metadata. It did not locate the exact integrated combination of per-scaffold review leases, semantic-staleness review, institutional dependency gates, accommodation exceptions, reversible access tiers, tombstones, and tested historical reconstruction. This is only a bounded search result: world novelty, patentability, freedom to operate, market size, and realized impact remain unmeasured.","arm":"COMPLETE_PROPOSAL_PORTFOLIO","candidate_version":0,"controller_recommendation":{"action":"STOP_EMPIRICAL_RESEARCH_NEEDED","repairable":false,"material_progress_observed":true,"progress_targets":["Secure a named module partner and written confirmation of the instructor, platform-operator, and accessibility authorities.","Quantify the baseline scaffold inventory and prevalence of stale, duplicated, unowned, indefinitely active, or reference-broken supports.","Preregister the three-condition matched-form sandbox study, including a minimum meaningful advantage and critical-error, accommodation, restoration, and workload falsifiers.","Validate the scaffold-dependence probe so that reduced support is not confounded with an unfamiliar interface or missing instruction.","Measure dependency-map completeness using declared references, runtime observations, and human review.","Demonstrate reversible state transitions and successful reconstruction of at least one prior support configuration without exposing sensitive learner or accommodation data.","Instrument reviewer time and engineering effort to replace assumption-driven cost bands with observed resource estimates."],"reason":"Bounded web research found substantial prior art and made the residual claim testable, but it cannot determine local problem prevalence, comparative benefit, dependency-map completeness, accommodation safety, review burden, or restoration reliability. Those questions require institutional data, field participation, and live sandbox testing, so further web search cannot clear the evidence gates."},"proposal_index":5}