{"id":"W4360978668","doi":"10.1145/3581754.3584136","title":"Supporting Qualitative Analysis with Large Language Models: Combining Codebook with GPT-3 for Deductive Coding","year":2023,"lang":"en","type":"preprint","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":207,"is_retracted":false,"has_abstract":true,"ca_institutions":"Microsoft (Canada); Research Canada","funders":"","keywords":"Codebook; Generalizability theory; Computer science; Coding (social sciences); Task (project management); Natural language processing; Curiosity; Qualitative analysis; Qualitative research; Task analysis; Artificial intelligence; Qualitative property; Machine learning; Data science; Psychology; Social psychology","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.0570225,0.0020826,0.001092118,0.004773885,0.001790627,0.006712057,0.005193062,0.002215771,0.01009599],"category_scores_gemma":[0.3246928,0.001458562,0.002278858,0.003599897,0.003625297,0.01032798,0.01293125,0.005140884,0.006454557],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.005164192,"about_ca_system_score_gemma":0.009166841,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.007324775,"about_ca_topic_score_gemma":0.01213574,"domain_scores_codex":[0.9397489,0.04490444,0.004260631,0.004840308,0.005573328,0.0006724051],"domain_scores_gemma":[0.6713411,0.2496595,0.008402717,0.04192705,0.02628511,0.00238464],"domain_codex":null,"domain_gemma":"methods","domain_candidate":"methods","domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0006918338,0.000399526,0.007821606,0.00343786,0.0002063834,0.0006778925,0.057498,0.0452274,0.01147138,0.1668606,0.03423222,0.6714752],"study_design_scores_gemma":[0.0002857561,0.0001656042,0.001099973,0.001303702,0.00006905326,0.0003541228,0.008000993,0.5375885,0.01160096,0.3886332,0.05066874,0.00022932],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.003922875,0.00006566569,0.9872041,0.0006891242,0.00006374536,0.0009120545,0.001547644,0.004160517,0.001434217],"genre_scores_gemma":[0.05407687,0.00005728307,0.9389969,0.0002467099,0.00002382089,0.002256502,0.003206886,0.0005752765,0.0005597575],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.9429775,"threshold_uncertainty_score":0.3015674,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.08960805608801989,"score_gpt":0.3901986999772469,"score_spread":0.3005906438892271,"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."}}