BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Strategic Classification - Jiri Hron (University of Cambridge)
DTSTART:20200415T100000Z
DTEND:20200415T113000Z
UID:TALK141589@talks.cam.ac.uk
CONTACT:75379
DESCRIPTION:When machine learning algorithms are deployed\, we often assum
 e their\npredictions do not affect the outcomes they are trained to foreca
 st.\nHowever\, this principle can break down when welfare (employment\,\ne
 ducation\, health) of rational individuals is affected. Knowing\ninformati
 on about the algorithm\, such individuals may manipulate their\nattributes
  in order to obtain a personally beneficial prediction (e.g.\,\na higher c
 redit score rating).\n\nThis reading group focuses on whether and how we c
 an design machine\nlearning algorithms which achieve non-trivial performan
 ce even in\npresence of strategic individuals. We further explore the pote
 ntially\nnegative social effects of strategic classification\, and the bro
 ader\nphenomenon of performativity of which strategic classification is a\
 nspecial case.
LOCATION:https://meet.google.com/hmo-aaio-agh
END:VEVENT
END:VCALENDAR