计算机视觉中的偏见-比面部识别更大.pdf

1、Bias in Computer Vision Its Bigger than Facial Recognition!Susan Kennedy,PhDAssistant Professor of PhilosophySanta Clara UniversityThe Optimistic View2 2023 Santa Clara UniversityBias in Computer Vision:Facial Recognition3 2023 Santa Clara UniversityBias in CV:Looking Beyond Facial Recognition4 2023

2、 Santa Clara UniversityAgriculturePlant Disease DetectionManufacturingQuality InspectionTransportationPothole DetectionBias can pose an ethical challenge,even without sensitive data!Bias in CV:Looking Beyond Facial Recognition5 2023 Santa Clara UniversityAgriculturePlant Disease DetectionManufacturi

3、ngQuality InspectionTransportationPothole Detection Expand the ethical circle to take into account the full range of stakeholdersWhodoes it not work for?Whatdoes it not work for?Whendoes it not work?Human SubjectsHuman Impacts6 2023 Santa Clara UniversityCoreDirectIndirect Toolkits Google What-If IB

4、M Fairness 360 Microsoft FairLearnMitigating Bias Technical Solutions7 2023 Santa Clara University Systematic errors stemming from bias in the datasets and algorithmic processes used Human bias present across the AI lifecycle and in the use of AI once deployed Present in the datasets used in AI,and

5、the institutional norms and practices across the AI lifecycle and in broader societyThe Tip of the Iceberg8 2023 Santa Clara University Bias-free AI is an unachievable goal Mitigating bias requires a bias-aware approach AI exists within a larger social system Mitigating bias requires a sociotechnica

6、l approachBias-Aware not Bias-Free!9 2023 Santa Clara University Instead of waiting for bias to strike,build a habit of anticipating preventable causes so they can be mitigated Across the entire AI lifecycle From pre-design to deployment 3 key problem areas:Datasets Testing and e