BEGIN:VCALENDAR VERSION:2.0 METHOD:PUBLISH PRODID:-//Telerik Inc.//Sitefinity CMS 15.1//EN BEGIN:VTIMEZONE TZID:Eastern Standard Time BEGIN:STANDARD DTSTART:20251102T020000 RRULE:FREQ=YEARLY;BYDAY=1SU;BYHOUR=2;BYMINUTE=0;BYMONTH=11 TZNAME:Eastern Standard Time TZOFFSETFROM:-0400 TZOFFSETTO:-0500 END:STANDARD BEGIN:DAYLIGHT DTSTART:20250301T020000 RRULE:FREQ=YEARLY;BYDAY=2SU;BYHOUR=2;BYMINUTE=0;BYMONTH=3 TZNAME:Eastern Daylight Time TZOFFSETFROM:-0500 TZOFFSETTO:-0400 END:DAYLIGHT END:VTIMEZONE BEGIN:VEVENT DESCRIPTION:Autonomous Systems Policy Institute (ASPI) speaker series prese nts Ferdinando Fioretto.Many agencies release statistics about groups of i ndividuals\nthat are then used as input to critical decision processes. Fo r example\, census\ndata is used to allocate funds and distribute \;&n bsp\;resources to states\nand jurisdictions. \;Similarly\, corporates are increasingly adopting \;machine\nlearning systems to derive socio- technical \;decisions\, including\ncriminal \;assessment\, landing \, and hiring. \;The resulting decisions can\nhave significant societa l and economic impacts for participating\nindividuals. \;\nIn many cas es\, the released data contain sensitive information whose privacy is\nstr ictly regulated and Differential Privacy has become the paradigm of\nchoic e for protecting data \;privacy. \;However\, while differential\np rivacy provides strong privacy guarantees\, it has become apparent recentl y\nthat it may induce biases and fairness issues in \;downstream decis ion\nprocesses\, including the allotment of federal funds\, apportionment of\ncongressional seats\, and learning biased classification results in le nding and\nhiring. These issues may adversely affect the health\, well-bei ng\, and sense of\nbelonging of many individuals\, and are currently poorl y understood. \;\n\nThis talk will describe our efforts in understandi ng and addressing these\nissues at the intersection of privacy\, fairness\ , and decision processes. I will\nfirst review the notion of Differential& nbsp\;Privacy and discuss its\napplications in data release and learning t asks. I will then examine the\nsocietal impacts of privacy under a fairnes s lens and shed light on what\naspects of the \;private algorithms and the data may be responsible in\nexacerbating unfairness. Finally\, I will propose a path to partially address\nthese fairness issues. The talk will conclude with an open discussion on the\nneed for tools that may be used by policymakers to test and address unfairness\nin privacy-preserving deci sion processes. \;Join Zoom Meeting\n\nhttps://syracuseuniversity.zoom .us/j/94181628976?pwd=ZUZPZHBRajAwMWJrbFRHb3g3Z25Vdz09Any questions\, plea se contact Lynnell Cabezas at \;lncabeza@syr.edu DTEND:20210929T200000Z DTSTAMP:20260317T125110Z DTSTART:20210929T190000Z LOCATION: SEQUENCE:0 SUMMARY:Privacy-Preserving Machine Learning: Uses and Unintended Disparate Effects UID:RFCALITEM639093342702484087 X-ALT-DESC;FMTTYPE=text/html:

Autonomous Systems Policy Institute (ASPI) speaker series presents Ferdinando Fioretto.


Many agencie s release statistics about groups of individuals\nthat are then used as in put to critical decision processes. For example\, census\ndata is used to allocate funds and distribute \; \;resources to states\nand jurisd ictions. \;Similarly\, corporates are increasingly adopting \;mach ine\nlearning systems to derive socio-technical \;decisions\, includin g\ncriminal \;assessment\, landing\, and hiring. \;The resulting d ecisions can\nhave significant societal and economic impacts for participa ting\nindividuals. \;
\nIn many cases\, the released data contain s ensitive information whose privacy is\nstrictly regulated and Different ial Privacy has become the paradigm of\nchoice for protecting data&nbs p\;privacy. \;However\, while differential\nprivacy provides strong pr ivacy guarantees\, it has become apparent recently\nthat it may induce bia ses and fairness issues in \;downstream decision\nprocesses\, includin g the allotment of federal funds\, apportionment of\ncongressional seats\, and learning biased classification results in lending and\nhiring. These issues may adversely affect the health\, well-being\, and sense of\nbelong ing of many individuals\, and are currently poorly understood. \;
\ n

\nThis talk will describe our efforts in understanding and add ressing these\nissues at the intersection of privacy\, fairness\, and deci sion processes. I will\nfirst review the notion of Differential \;Priv acy and discuss its\napplications in data release and learning tasks. I wi ll then examine the\nsocietal impacts of privacy under a fairness lens and shed light on what\naspects of the \;private algorithms and the data may be responsible in\nexacerbating unfairness. Finally\, I will propose a path to partially address\nthese fairness issues. The talk will conclude with an open discussion on the\nneed for tools that may be used by policym akers to test and address unfairness\nin privacy-preserving decision proce sses. \;


Join Zoom Meeting

\n\n

https://syracuseuniversity.zoom.us/j/94181628976?pwd=ZUZPZHBRajAwMWJ rbFRHb3g3Z25Vdz09


Any questions\, please contact Lynn ell Cabezas at \;lncabeza@syr.edu

END:VEVENT END:VCALENDAR