BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Fairness in AI - Elre Oldewage (University of Cambridge)
DTSTART:20210217T110000Z
DTEND:20210217T123000Z
UID:TALK156727@talks.cam.ac.uk
CONTACT:Elre Oldewage
DESCRIPTION:Applications of machine learning almost always involve data ab
 out people. Unfortunately\, training data for such applications usually en
 code the demographic disparities that are present (or have historically be
 en present) in our society. This\, in turn\, may lead to machine learning 
 solutions that re-enforce discrimination -- not necessarily on purpose\, b
 ut due to bias in the training data. As machine learning researchers\, the
  solution may seem simple: get better\, unbiased data. However\, as we wil
 l discuss\, bias is nigh impossible to escape completely. Instead\, we mus
 t develop techniques to identify and counter biased decision-making by ML 
 models.\n\nIn this talk\, we will provide motivational examples and discus
 s a number of sources from which bias and discrimination may arise. We wil
 l consider different definitions for fairness\, eventually boiling them do
 wn to three main criteria: independence\, separation and sufficiency. We'l
 l also discuss a number of bias mitigation strategies\, by which these cri
 teria may be achieved. Finally\, we'll discuss the limitations of the thre
 e observational criteria and briefly discuss fairness from a causal perspe
 ctive.\n\nRecommended reading:\n\nMehrabi\, N.\, Morstatter\, F.\, Saxena\
 , N.\, Lerman\, K.\, and Galstyan\, A. _A Survey on Bias and Fairness in M
 achine Learning._ arXiv e-prints\, art. arXiv:1908.09635\, 2019. Available
  at: https://arxiv.org/abs/1908.09635\n\nAdditional reading\, for those in
 terested in the topic\, is the Fair ML Book\, available for free online:\n
 \nBarocas\, S.\, Hardt\, M.\, and Narayanan\, A. _Fairness and Machine Lea
 rning._ fairmlbook.org\, 2019. Available at: http://www.fairmlbook.org
LOCATION:https://eng-cam.zoom.us/j/86068703738?pwd=YnFleXFQOE1qR1h6Vmtwbno
 0LzFHdz09
END:VEVENT
END:VCALENDAR