{"id":"W2333968622","doi":"10.1177/154193120204601408","title":"How Good is Search Engine Ranking? a Validation Study with Human Judges","year":2002,"lang":"en","type":"article","venue":"Proceedings of the Human Factors and Ergonomics Society Annual Meeting","topic":"Information Retrieval and Search Behavior","field":"Computer Science","cited_by":3,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto","funders":"","keywords":"Ranking (information retrieval); Relevance (law); Information retrieval; Search engine; Set (abstract data type); Correlation; Psychology; Computer science; Contrast (vision); Statistics; Mathematics; Artificial intelligence; Political science; Law","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":[],"consensus_categories":[],"category_scores_codex":[0.0005487966,0.0001831531,0.0002090247,0.00005328889,0.0008947063,0.0005643988,0.0006306058,0.00006499788,0.000006197114],"category_scores_gemma":[0.0000202509,0.0001212261,0.0001154161,0.000276145,0.0001078189,0.001210655,0.0003980236,0.0002588471,7.086069e-7],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00005804358,"about_ca_system_score_gemma":0.000009887454,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00003861677,"about_ca_topic_score_gemma":0.000001657791,"domain_scores_codex":[0.9987895,0.0000108599,0.0002533369,0.000271829,0.0003844835,0.0002899153],"domain_scores_gemma":[0.9991997,0.00003313162,0.000224742,0.0001346814,0.0003342204,0.00007345119],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.00002189014,0.0004911838,0.686015,0.0002625038,0.000193647,2.69192e-7,0.2929847,0.00004039289,0.009616455,0.007249291,0.001315697,0.001809065],"study_design_scores_gemma":[0.005793343,0.003797684,0.534888,0.0004257084,0.0002563106,0.00002012078,0.19215,0.02212882,0.2362149,0.0007583387,0.001473102,0.002093771],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9987661,0.00001742083,0.00008251624,0.0003577446,0.00003206701,0.0003553533,0.000007906501,0.00005744445,0.0003235073],"genre_scores_gemma":[0.9985055,0.000008716303,0.0009771145,0.00005240781,0.00004335372,0.000007781991,0.000001275562,0.00001136425,0.000392498],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.2265984,"threshold_uncertainty_score":0.6881448,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02802410079066921,"score_gpt":0.2392387341369389,"score_spread":0.2112146333462697,"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."}}