BEGIN:VCALENDAR
VERSION:2.0
CALSCALE:GREGORIAN
PRODID:iCalendar-Ruby
BEGIN:VEVENT
CATEGORIES:
DESCRIPTION:Speaker: Aaron Roth (University of Pennsylvania)\n\nTitle: Trac
 table Agreement Protocols\n\nAbstract: As ML models become increasingly pow
 erful\, it is an attractive proposition to use them in important decision m
 aking pipelines\, in collaboration with human decision makers. But how shou
 ld a human being and a machine learning model collaborate to reach decision
 s that are better than either of them could achieve on their own? If the hu
 man and the ML model were perfect Bayesians\, operating in a setting with a
  commonly known and correctly specified prior\, Aumann's classical agreemen
 t theorem would give us one answer: they could engage in conversation about
  the task at hand\, and their conversation would be guaranteed to converge 
 to (accuracy-improving) agreement. This classical result however would requ
 ire making many implausible assumptions\, both about the knowledge and comp
 utational power of both parties. We show how to recover similar (and more g
 eneral) results using only computationally and statistically tractable assu
 mptions\, which substantially relax full Bayesian rationality. We further g
 ive weak-learning conditions under which this collaboration will result in 
 "information aggregation" --- i.e. predictions that are as accurate as coul
 d have been made by a model that had access to -both- party's observations\
 , even though neither party in the interaction actually has access to these
  pooled observations. \n\n \n\nJoint work with Natalie Collina\, Varun Gupt
 a\, and Surbhi Goel\, based on a paper that will appear in STOC 2025\, and 
 with Natalie Collina\, Ira Globus-Harris\, Varun Gupta\, Surbhi Goel\, and 
 Mirah Shi based on a new preprint. \n\n \n\nBiography: Aaron Roth is the He
 nry Salvatori Professor of Computer and Cognitive Science\, in the Computer
  and Information Sciences department at the University of Pennsylvania\, wi
 th a secondary appointment in the Wharton statistics department. He is affi
 liated with the Warren Center for Network and Data Science\, and co-directo
 r of the Networked and Social Systems Engineering (NETS) program.  He is al
 so an Amazon Scholar at Amazon AWS. He is the recipient of the Hans Sigrist
  Prize\, a Presidential Early Career Award for Scientists and Engineers (PE
 CASE)\, an Alfred P. Sloan Research Fellowship\, an NSF CAREER award\, and 
 research awards from Yahoo\, Amazon\, and Google.  His research focuses on 
 the algorithmic foundations of data privacy\, algorithmic fairness\, game t
 heory\, learning theory\, and machine learning.  Together with Cynthia Dwor
 k\, he is the author of the book “The Algorithmic Foundations of Differenti
 al Privacy.” Together with Michael Kearns\, he is the author of “The Ethica
 l Algorithm”.
DTEND:20250502T160000Z
DTSTAMP:20260310T214945Z
DTSTART:20250502T150000Z
GEO:42.362019;-71.087844
LOCATION:Building E18\, 304
SEQUENCE:0
SUMMARY:Stochastics and Statistics Seminar
UID:tag:localist.com\,2008:EventInstance_48844672028436
URL:https://calendar.mit.edu/event/stochastics-and-statistics-seminar-9539
END:VEVENT
END:VCALENDAR