{"paper":{"title":"GESD: Beyond Outcome-Oriented Fairness","license":"http://creativecommons.org/licenses/by/4.0/","headline":"GESD measures fairness by tracking how consistently machine learning models explain their predictions across demographic subgroups.","cross_cats":["cs.AI","cs.CY"],"primary_cat":"cs.LG","authors_text":"Gideon Popoola, John Sheppard","submitted_at":"2026-05-14T18:10:57Z","abstract_excerpt":"Machine learning (ML) algorithms are increasingly deployed in high-stakes decision-making domains such as loan approvals, hiring, and recidivism predictions. While existing fairness metrics (e.g., statistical parity, equal opportunity) effectively quantify outcome-oriented disparities, they offer limited insight into the procedure or explanation behind biased decisions. To address this gap, we propose Group-level Explanation Stability Disparity (GESD), a \\textit{procedural-oriented} fairness metric that measures disparities in the stability, robustness, and sensitivity of model explanations ac"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"GESD effectively captures group-wise discrepancies in explanation quality, and that FEU improves both utility and fairness over state-of-the-art methods.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That measuring disparities in explanation stability, robustness, and sensitivity across subgroups provides a meaningful and model-agnostic indicator of procedural fairness without requiring additional assumptions about the underlying explainer or data distribution.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"The paper proposes GESD, a procedural fairness metric for group disparities in explanation stability and robustness, and integrates it into the FEU multi-objective optimization framework.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"GESD measures fairness by tracking how consistently machine learning models explain their predictions across demographic subgroups.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"1ca1dc8307a2f54bc4e5b4dab6073b2f5d260198e71334d69e3ef06f5cdd119b"},"source":{"id":"2605.15295","kind":"arxiv","version":1},"verdict":{"id":"0cd32de7-fc37-47ae-8460-3f4d0d03193e","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T16:26:09.804352Z","strongest_claim":"GESD effectively captures group-wise discrepancies in explanation quality, and that FEU improves both utility and fairness over state-of-the-art methods.","one_line_summary":"The paper proposes GESD, a procedural fairness metric for group disparities in explanation stability and robustness, and integrates it into the FEU multi-objective optimization framework.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That measuring disparities in explanation stability, robustness, and sensitivity across subgroups provides a meaningful and model-agnostic indicator of procedural fairness without requiring additional assumptions about the underlying explainer or data distribution.","pith_extraction_headline":"GESD measures fairness by tracking how consistently machine learning models explain their predictions across demographic subgroups."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.15295/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_compliance","ran_at":"2026-05-19T16:36:04.312407Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-19T16:31:18.364619Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-19T14:41:54.234657Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T13:33:22.783854Z","status":"skipped","version":"1.0.0","findings_count":0}],"snapshot_sha256":"a98ec30a092bbb9db1e30665e9f52baf73556af5ee2dcf02d682746a59cbc5fc"},"references":{"count":29,"sample":[{"doi":"","year":2016,"title":"Big data’s disparate impact,","work_id":"8ef3fb2b-65b4-43d7-8193-87036d2fa1ed","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"Investigating and mitigating the performance–fairness tradeoff via protected-category sampling,","work_id":"f9665236-7427-4074-b52c-1cdb2354e670","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2017,"title":"Optimized pre-processing for discrimination prevention","work_id":"d285c33d-8270-4df5-a8ee-82dc61103423","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2017,"title":"Fair prediction with disparate impact: A study of bias in recidivism prediction instruments,","work_id":"7a40f3a2-2b53-4d63-bed0-b01fd10cf685","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2016,"title":"Equality of opportunity in super- vised learning,","work_id":"333f5ab2-14d7-4d29-a133-71e893194478","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":29,"snapshot_sha256":"c2753c2a08086e38f4cb6240f458c30266967cba1eb89159d91fdb882f4493f3","internal_anchors":1},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}