{"id":"W7105683204","doi":"10.1109/asew67777.2025.00035","title":"ForeSPECT: A Model-Driven Framework for Validation and Traceability in Forecasting Systems","year":2025,"lang":"","type":"article","venue":"","topic":"Forecasting Techniques and Applications","field":"Decision Sciences","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Royal Canadian Navy","funders":"NAV","keywords":"Traceability; Metamodeling; TRACE (psycholinguistics); Representation (politics); Domain (mathematical analysis); Foundation (evidence)","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.004049278,0.0002912893,0.0005717078,0.0005104378,0.000464701,0.0007807559,0.0006354876,0.0003957946,0.0000301673],"category_scores_gemma":[0.00525879,0.0002522219,0.0001564838,0.001687953,0.0002162861,0.0003695292,0.0002844897,0.0003311774,0.000003658743],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.000174905,"about_ca_system_score_gemma":0.000230454,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001784349,"about_ca_topic_score_gemma":0.0001394659,"domain_scores_codex":[0.9962065,0.0001325166,0.001569016,0.001137919,0.0004596403,0.0004944085],"domain_scores_gemma":[0.9933563,0.004802096,0.0003897976,0.0008483243,0.000488623,0.0001148923],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00009072461,0.0001706214,0.006148473,0.0001770639,0.00001378511,3.325948e-7,0.0006388321,0.04849026,0.00003911968,0.8956887,0.00165894,0.04688311],"study_design_scores_gemma":[0.0002092105,0.00005440053,0.0002160543,0.0003037051,0.0000176714,0.000001687998,0.0004604307,0.564372,0.0001353409,0.433624,0.0004726869,0.0001328531],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1025969,0.0002569838,0.8870451,0.001866164,0.0001999545,0.003046337,0.00008195401,0.0001463231,0.004760191],"genre_scores_gemma":[0.7154939,0.00002626373,0.2827109,0.00005717503,0.0000363708,0.0006008468,0.000004718554,0.00001285507,0.001056913],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.612897,"threshold_uncertainty_score":0.999993,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.2125855422147073,"score_gpt":0.4343277666536215,"score_spread":0.2217422244389142,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}