{"id":"W4415195142","doi":"10.1609/aies.v8i3.36706","title":"Measuring What Matters: Connecting AI Ethics Evaluations to System Attributes, Hazards, and Harms","year":2025,"lang":"en","type":"article","venue":"Proceedings of the AAAI/ACM Conference on AI Ethics and Society","topic":"Ethics and Social Impacts of AI","field":"Social Sciences","cited_by":3,"is_retracted":false,"has_abstract":true,"ca_institutions":"McGill University","funders":"","keywords":"Harm; Framing (construction); Situated; Set (abstract data type); Focus (optics)","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":[0.2283404,0.001531562,0.001813081,0.02147307,0.004440017,0.01842444,0.002265716,0.003026392,0.003051084],"category_scores_gemma":[0.5520977,0.001156204,0.001662066,0.0121426,0.02004141,0.03290143,0.01223895,0.005780597,0.0005433692],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.01283422,"about_ca_system_score_gemma":0.01372288,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.004130338,"about_ca_topic_score_gemma":0.004801297,"domain_scores_codex":[0.6932099,0.2428885,0.01906711,0.006422298,0.03515738,0.003254858],"domain_scores_gemma":[0.25746,0.6085314,0.05058457,0.01960569,0.06038386,0.003434563],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.0005750891,0.0004249106,0.1196097,0.01072254,0.001335346,0.0001730054,0.06621506,0.009404719,0.0008872233,0.3233348,0.005530156,0.4617874],"study_design_scores_gemma":[0.0001073307,0.000820675,0.06757854,0.01586709,0.001187042,0.0002565269,0.05940308,0.01477449,0.004205769,0.7908171,0.04458164,0.0004007155],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.3464628,0.02723847,0.4347551,0.05734527,0.0009856529,0.002295407,0.001214612,0.000574714,0.1291279],"genre_scores_gemma":[0.932776,0.002883302,0.06007252,0.001899654,0.0001546456,0.001159491,0.0002307115,0.0001222223,0.0007012986],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.2283404,"threshold_uncertainty_score":0.9515937,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1888758223803678,"score_gpt":0.4198797978285312,"score_spread":0.2310039754481635,"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."}}