{"id":"W4415165924","doi":"10.1007/s00146-025-02670-7","title":"Not all AI-generated faces are created equal: impacts of model gender, race, and emotional expression on classification accuracy","year":2025,"lang":"en","type":"article","venue":"AI & Society","topic":"Ethics and Social Impacts of AI","field":"Social Sciences","cited_by":0,"is_retracted":false,"has_abstract":false,"ca_institutions":"University of Ottawa","funders":"","keywords":"Facial expression; Emotional expression; Generative grammar; Emotional intelligence; Face (sociological concept); Expression (computer science); Variance (accounting); Generative model; Emotion recognition","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":[],"consensus_categories":[],"category_scores_codex":[0.01510366,0.0007860262,0.0007631428,0.0004363903,0.001102854,0.003900747,0.000962098,0.001307502,0.004584179],"category_scores_gemma":[0.09765639,0.0003943477,0.0006477593,0.0003168572,0.001115918,0.003714812,0.001809422,0.002440954,0.002352364],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0008095028,"about_ca_system_score_gemma":0.0007621551,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.005564772,"about_ca_topic_score_gemma":0.005396622,"domain_scores_codex":[0.9939551,0.003580369,0.000274981,0.0008378857,0.000881802,0.0004698799],"domain_scores_gemma":[0.9156393,0.06977205,0.002065509,0.00696725,0.004521259,0.001034755],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"observational","study_design_scores_codex":[0.02169657,0.0016638,0.3376351,0.0004467261,0.001574439,0.000635785,0.003901406,0.08980111,0.03611138,0.007360119,0.02511737,0.4740562],"study_design_scores_gemma":[0.0003306269,0.002002987,0.2155995,0.00038557,0.001103765,0.001635291,0.005844874,0.7010683,0.0330194,0.02874098,0.01001201,0.0002566709],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9669567,0.0006354535,0.01529791,0.002347536,0.0006562418,0.0001077266,0.0006681663,0.0004502402,0.01287989],"genre_scores_gemma":[0.9926524,0.0001056262,0.00376274,0.0003914723,0.00006827378,0.00003501098,0.0006809321,0.0001730531,0.002130442],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01510366,"threshold_uncertainty_score":0.07987678,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.14485044819305,"score_gpt":0.4251979108399317,"score_spread":0.2803474626468817,"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."}}