{"id":"W4389161391","doi":"10.2139/ssrn.4617116","title":"Using Large Language Models to Support Thematic Analysis in Empirical Legal Studies","year":2023,"lang":"en","type":"article","venue":"SSRN Electronic Journal","topic":"Artificial Intelligence in Law","field":"Social Sciences","cited_by":24,"is_retracted":false,"has_abstract":false,"ca_institutions":"Université de Montréal","funders":"","keywords":"Thematic analysis; Thematic map; Coding (social sciences); Computer science; Qualitative analysis; Empirical research; Quality (philosophy); Qualitative research; Data science; Natural language processing; Sociology; Epistemology; Social science; Geography; Cartography","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":[0.006704684,0.0001253515,0.0003349342,0.0006184372,0.0005525195,0.00009958541,0.0003827228,0.00007599148,0.00009685264],"category_scores_gemma":[0.0004250585,0.0001116457,0.000186575,0.003266528,0.0001052116,0.0004029915,0.00008644312,0.0007990508,0.0001468965],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001547356,"about_ca_system_score_gemma":0.001626901,"about_ca_topic_candidate":true,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001653489,"about_ca_topic_score_gemma":0.1609637,"domain_scores_codex":[0.9959061,0.0003065108,0.0004274638,0.0002263673,0.0005828259,0.002550779],"domain_scores_gemma":[0.9993236,0.0001816333,0.0001022541,0.0001532436,0.0001106453,0.000128628],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"qualitative","study_design_scores_codex":[0.000046703,0.0001798394,0.0117613,0.0000103913,0.001200836,0.0001641371,0.2648571,0.04610487,0.000234319,0.6673884,0.000502278,0.007549759],"study_design_scores_gemma":[0.0001210255,0.0001345811,0.00009341377,0.00002978858,0.0002417443,0.00001886151,0.6285772,0.01880708,0.0000799843,0.3507158,0.0008772149,0.0003033039],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9772111,0.0004316067,0.01807781,0.001810756,0.0001662374,0.0001701826,0.000002727065,0.00007310647,0.002056473],"genre_scores_gemma":[0.9968623,0.0006047161,0.00015551,0.0001994884,0.0002487446,0.000006278062,0.00000140672,0.00001494368,0.001906607],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.3637201,"threshold_uncertainty_score":0.8543465,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1935123808274758,"score_gpt":0.4858269344481921,"score_spread":0.2923145536207162,"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."}}