BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:I’m not good at this\, so don’t match me
DTSTART:20261120T111500
DTEND:20261120T120000
DTSTAMP:20261109T143855Z
UID:805ec766f47172e0866f1b0fda1cb2d74f4ffa5ff0bf0535840a18d4
CATEGORIES:Conferences - Seminars
DESCRIPTION:Dr Stefania Gavrila-Ionescu\, postdoctoral researcher at the 
 Automatic Control Laboratory (ETH Zurich)\nAbstract:\nIn two-sided matchin
 g markets such as school choice and refugee assignment\, it is difficult t
 o pair agents optimally. To aid\, (1) machine learning models can make pre
 dictions of future outcomes based on historical data\, which are (2) used 
 to produce more efficient matchings. Although deployed together\, these pr
 ediction and matching mechanisms are usually analyzed separately. But what
  happens if we consider the market as a whole? In this talk\, I present ad
 versarial interaction attacks: a strategic behavior of institutions that c
 an adapt their actions to create targeted predictions in the future. This 
 strategic behavior is not merely an attack on the reported data but a chan
 ge in the actual interaction between agents. I will discuss when such atta
 cks are optimal\, their consequences (e.g.\, inefficiencies and unfairness
 )\, and mitigation strategies. Ultimately\, the school choice and refugee 
 assignment examples call for analyzing shifts in strategies before deployi
 ng AI systems\, as they can have severe societal consequences.\n\nBiograph
 y:\nStefania Gavrila-Ionescu is a postdoctoral researcher at the Automatic
  Control Laboratory (ETH Zurich) working on responsible automation. She ha
 s a background in Mathematics (BA from the University of Cambridge)\, Logi
 c\, Language\, and Computation (MSc with distinction from the University o
 f Amsterdam and ILLC)\, and Computer Science (PhD with distinction from th
 e University of Zurich\, Social Computing). She investigates the role of f
 ilter\, search\, sort\, moderation\, and recommendation tools in shaping o
 ur society. In doing so\, she aims to both understand the effects of these
  tools on society and find what methods are generally effective in doing s
 o. As such\, her prior work spreads across various application domains (e.
 g.\, school choice\, foster care\, online dating\, content-creator-centere
 d platforms) and different areas of study (e.g.\, Algorithmic Fairness\, R
 ecommender Systems\, AI\, Social Networks\, Game Theory\, Matching Markets
 ). From a methodological perspective\, she primarily uses simulations and 
 theoretical analysis.
LOCATION:ME C2 405 https://plan.epfl.ch/?room==ME%20C2%20405
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR