{"paper":{"title":"Auditing Demographic Bias in Facial Landmark Detection for Fair Human-Robot Interaction","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"After controlling for head pose and image resolution, facial landmark detectors show no bias by gender or race but retain an age-related bias.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Jos\\'e M. Buenaposada, Luis Baumela, Pablo Parte, Roberto Valle","submitted_at":"2026-04-08T11:22:18Z","abstract_excerpt":"Fairness in human-robot interaction critically depends on the reliability of the perceptual models that enable robots to interpret human behavior. While demographic biases have been widely studied in high-level facial analysis tasks, their presence in facial landmark detection remains unexplored. In this paper, we conduct a systematic audit of demographic bias in this task, analyzing the age, gender, and race biases. To this end, we introduce a controlled statistical methodology to disentangle demographic effects from confounding visual factors. Our analysis demonstrates that visual confounder"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Notably, after accounting for these confounders, we show that performance disparities across gender and race vanish. However, we identify a statistically significant age-related effect, with higher biases observed for older individuals.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The controlled statistical methodology successfully disentangles demographic effects from confounding visual factors such as head pose and image resolution without residual interactions or dataset-specific artifacts that could mask or create apparent biases.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"After accounting for confounding factors like head pose and resolution, demographic biases in facial landmark detection largely vanish except for a statistically significant age effect disadvantaging older people.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"After controlling for head pose and image resolution, facial landmark detectors show no bias by gender or race but retain an age-related bias.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"593f59561054f888fee82e9e42a799d236b9e6a723ec0e1bdf151c5a389afdbe"},"source":{"id":"2604.06961","kind":"arxiv","version":2},"verdict":{"id":"b5d7b65c-873f-4ea3-a7b3-0bedc8e9068c","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T18:02:22.400510Z","strongest_claim":"Notably, after accounting for these confounders, we show that performance disparities across gender and race vanish. However, we identify a statistically significant age-related effect, with higher biases observed for older individuals.","one_line_summary":"After accounting for confounding factors like head pose and resolution, demographic biases in facial landmark detection largely vanish except for a statistically significant age effect disadvantaging older people.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The controlled statistical methodology successfully disentangles demographic effects from confounding visual factors such as head pose and image resolution without residual interactions or dataset-specific artifacts that could mask or create apparent biases.","pith_extraction_headline":"After controlling for head pose and image resolution, facial landmark detectors show no bias by gender or race but retain an age-related bias."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.06961/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"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"}