{"id":"W3174398179","doi":"10.1145/3450614.3464619","title":"Goal Modeling-based Evaluation for Personalized Recommendation Systems","year":2021,"lang":"en","type":"article","venue":"","topic":"Recommender Systems and Techniques","field":"Computer Science","cited_by":6,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Ottawa","funders":"","keywords":"Computer science; Recommender system; Human–computer interaction; World Wide Web","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":[0.001199582,0.00009707447,0.0001493297,0.00006379705,0.0001090989,0.0002746741,0.0002091506,0.00006250518,0.00005265484],"category_scores_gemma":[0.00005676324,0.00008788586,0.00008643042,0.0001661568,0.000004554918,0.0003074947,0.00004006112,0.00004290269,0.000006483079],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001070479,"about_ca_system_score_gemma":0.0002102314,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00007240799,"about_ca_topic_score_gemma":0.00001646791,"domain_scores_codex":[0.9988097,0.0001836145,0.0002705991,0.000340038,0.0002381604,0.0001579128],"domain_scores_gemma":[0.9988949,0.00008480027,0.00007632694,0.0003181855,0.0005795558,0.00004621679],"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":[0.000020269,0.0002311714,0.0001227542,0.0002109526,0.0000783225,0.000002377303,0.0004217492,0.01344457,0.00156123,0.7960098,0.02865964,0.1592372],"study_design_scores_gemma":[0.0005950645,0.0000378609,0.000003017759,0.00002353629,0.000008184777,0.000006239565,0.00006412771,0.9679453,0.001413762,0.002177984,0.02760874,0.0001161975],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.0003081281,0.000174133,0.9906164,0.002828653,0.0006956921,0.0005929984,0.000003875864,0.0002784242,0.004501712],"genre_scores_gemma":[0.8396828,0.000006179105,0.1580832,0.0005366258,0.0001181181,0.0006269627,0.0001196596,0.00001250461,0.0008139633],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9545007,"threshold_uncertainty_score":0.3583881,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.07737232726247448,"score_gpt":0.3523079213138741,"score_spread":0.2749355940513996,"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."}}