{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp09_archetype_breadth150_20260804","research_id":"eoa_inverse_innovation_exp09_light_prior_art_20260804","cell_id":"prediction_error_learning_calibration__chemistry_materials","search_lanes":{"direct_problem_and_intervention":{"queries":["\"solid-state electrolyte\" screening \"prediction error\" conductivity","\"solid-state electrolyte\" Bayesian optimization conductivity screening","materials discovery confidence-adjusted surprise active learning"],"source_ids":["SRC1","SRC2"],"no_result_note":null},"synonyms_and_historical_terms":{"queries":["materials discovery active learning prediction residual experiment model calibration","materials science Bayesian surprise sequential learning","structure-property active learning model calibration"],"source_ids":["SRC1","SRC4"],"no_result_note":null},"products_practices_and_standards":{"queries":["solid electrolyte ionic conductivity measurement best practices pellet density impedance spectroscopy","solid electrolyte conductivity common standard frequency temperature reproducibility","NIST active learning materials structure property mapping"],"source_ids":["SRC3","SRC4"],"no_result_note":null},"component_combination":{"queries":["Bayesian optimization solid electrolyte prediction uncertainty noisy measurements","materials active learning surprising observation local verification model update","solid electrolyte conductivity physical parameters frequency density temperature"],"source_ids":["SRC1","SRC2","SRC3","SRC4"],"no_result_note":null}},"sources":[{"source_id":"SRC1","title":"Confidence Adjusted Surprise Measure for Active Resourceful Trials (CA-SMART): A Data-driven Active Learning Framework for Accelerating Material Discovery under Resource Constraints","publisher":"arXiv / Cornell University","url":"https://arxiv.org/abs/2503.21095","source_type":"PRIMARY_RESEARCH","claims_supported":["CA-SMART explicitly defines surprise as divergence between expected and observed material-property outcomes.","It weights surprise by predictive confidence, discounts deviations in uncertain regions, verifies surprising observations using nearby samples, and changes subsequent sampling behavior.","It reports comparisons with conventional Bayesian-optimization and surprise-based methods on synthetic functions and steel fatigue-strength prediction."]},{"source_id":"SRC2","title":"Accelerated discovery of solid-state electrolytes using Bayesian optimisation","publisher":"ChemRxiv","url":"https://chemrxiv.org/engage/api-gateway/chemrxiv/assets/orp/resource/item/670f0ae3cec5d6c142449f0c/original/accelerated-discovery-of-solid-state-electrolytes-using-bayesian-optimisation.pdf","source_type":"PRIMARY_RESEARCH","claims_supported":["Bayesian optimization has already been applied directly to solid-state-electrolyte discovery using a Gaussian-process surrogate that produces predictions and uncertainties.","Simulation outcomes are iteratively added to the training data to update the model and obtain new suggestions.","The workflow treats observation noise statistically and evaluates conductivity-related performance alongside bandgap and interfacial-stability requirements."]},{"source_id":"SRC3","title":"Reassessing the bulk ionic conductivity of solid-state electrolytes","publisher":"Royal Society of Chemistry","url":"https://pubs.rsc.org/en/content/articlehtml/2018/se/c8se00139a","source_type":"PRIMARY_RESEARCH","claims_supported":["Measured solid-electrolyte ionic conductivity depends on material properties and physical parameters including relative density, temperature, relaxation time, frequency, pellet dimensions, and impedance interpretation.","The study demonstrates large conductivity changes with impedance high-frequency limits and reports that measurement choices impede comparability.","The authors call for a common solid-state-electrolyte conductivity measurement standard."]},{"source_id":"SRC4","title":"Active learning for regression of structure-property mapping: the importance of sampling and representation","publisher":"National Institute of Standards and Technology","url":"https://www.nist.gov/publications/active-learning-regression-structure-property-mapping-importance-sampling-and","source_type":"PRIMARY_RESEARCH","claims_supported":["Active-learning regression and iterative enlargement of a materials data pool are established approaches for calibrating structure-property models.","Sampling strategy and representation affect the amount of evaluation data needed for robust material-property prediction.","The reported case studies provide an official-source example of model calibration and adaptive sampling in materials science."]}],"problem_evidence":{"status":"PARTLY_SUPPORTED","finding":"The general problem is visible: solid-electrolyte conductivity measurements can vary materially with pellet and protocol conditions, while Bayesian and active-learning workflows repeatedly convert predictions and new outcomes into subsequent model or sampling decisions. This creates a credible risk that poorly controlled outcomes teach the wrong relation. The stronger site-specific premise—that real formulation campaigns commonly promote candidates by raw room-temperature conductivity without immutable predictions or expectation-adjusted review—was not directly demonstrated by the retained sources.","source_ids":["SRC2","SRC3","SRC4"]},"closest_prior_art":[{"name":"CA-SMART confidence-adjusted surprise learning","source_ids":["SRC1"],"overlap":"Directly uses prediction-versus-outcome surprise, predictive confidence, verification of surprising observations, and surprise-dependent changes to subsequent material experiments. This collides with the proposal's core idea that reliable unexpected outcomes should drive more learning than expected outcomes.","remaining_difference":"CA-SMART is not specific to solid-state-electrolyte metrology and does not provide the proposal's full batch record, reference-pellet and instrument-drift filters, causal attribution trace to named composition or processing assumptions, personnel safeguards, or independent safety gates."},{"name":"Bayesian optimization for solid-state-electrolyte discovery","source_ids":["SRC2"],"overlap":"Uses prospective Gaussian-process predictions and uncertainties, evaluates new electrolyte candidates, models observation noise, incorporates results into training data, and iteratively produces new suggestions while separately checking stability-related properties.","remaining_difference":"The published workflow updates through ordinary Bayesian ingestion and acquisition rather than requiring an explicit signed residual, reproducibility and metrology eligibility decision, causal credit assignment, and residual-dependent update gain for every batch."},{"name":"Controlled solid-electrolyte conductivity measurement","source_ids":["SRC3"],"overlap":"Recognizes that conductivity can be distorted by pellet and measurement parameters and that comparable protocols are necessary before attributing performance to the material.","remaining_difference":"It improves measurement validity but does not specify how expectation-relative residuals should control scientific-model or experimental-queue updates."},{"name":"Active learning for materials structure-property calibration","source_ids":["SRC4"],"overlap":"Iteratively selects evaluations and calibrates predictive structure-property models, establishing the broader adaptive-learning practice.","remaining_difference":"It does not use signed surprise, metrology-based residual eligibility, or explicit feature-level credit assignment as the update rule."}],"prior_art_disposition":"SUBSTANTIAL_COLLISION","contrastive_claim_remaining":"The remaining testable distinction is the electrolyte-specific operational combination: prospectively freeze a protocol-conditioned conductivity prediction for every physical batch, preserve the residual's sign, determine eligibility using replicates, reference-pellet drift, impedance diagnostics and process context, assign the deviation to a named composition, process or metrology assumption, and bound the corresponding model or queue update without weakening stability or safety gates. The core surprise-weighted active-learning principle itself is already disclosed by CA-SMART.","contrastive_claim_falsifier":"The remaining distinction would be falsified by an existing electrolyte or comparable physical-materials workflow that combines prospective batch predictions, signed residuals, replicate or local verification, metrology and context filtering, named causal or feature-level credit assignment, and update magnitude or experimental priority explicitly conditioned on reliable surprise. Evidence that an existing laboratory's optimizer already performs these steps would reduce the proposal to documentation or terminology changes.","gates":{"adequate_source_search":{"status":"PASS","rationale":"The bounded search covered the proposal directly, older and synonymous active-learning terminology, electrolyte-screening and measurement practices, and combinations of surprise, uncertainty, verification, calibration and metrology. Exactly four opened sources from four publisher contexts were retained, all reporting primary research.","source_ids":["SRC1","SRC2","SRC3","SRC4"]},"supported_problem":{"status":"PASS","rationale":"The problem is partly supported: conductivity is protocol-sensitive and adaptive materials workflows use measured outcomes to update models and sampling. The claimed prevalence of raw-rank queue decisions remains unverified and should be tested locally.","source_ids":["SRC2","SRC3","SRC4"]},"distinct_testable_claim":{"status":"PASS","rationale":"Despite substantial collision on surprise-weighted learning, the remaining electrolyte-specific claim has observable records, eligibility criteria, attribution fields and update decisions that can be compared with ordinary Bayesian ingestion.","source_ids":["SRC1","SRC2","SRC3"]},"bounded_next_test":{"status":"PASS","rationale":"A prospective 24-batch shadow study is bounded, reversible and capable of comparing residual-gated simulated decisions with the existing workflow using one-step-ahead calibration, repeated-surprise reproducibility, attribution agreement and unchanged constraint violations.","source_ids":["SRC1","SRC2","SRC3","SRC4"]},"no_obvious_safety_or_authority_stop":{"status":"PASS","rationale":"Shadow operation on already approved batches avoids autonomous recipe or queue changes. Separate evaluation of stability and other electrolyte requirements is consistent with prior practice, while standardized measurement controls reduce rather than add laboratory risk. Independent safety vetoes must remain outside the learning objective.","source_ids":["SRC2","SRC3"]}},"screen_survival":false,"world_novelty_boundary":"This bounded public-web screen identifies substantial collision with surprise-driven, confidence-adjusted materials active learning and adjacent electrolyte Bayesian optimization and metrology practices. It does not establish world novelty, patentability, freedom to operate, market size, expert acceptance or realized value. Patents, proprietary laboratory systems, optimizer implementations and unindexed domain workflows were not exhausted."}