BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//NextToU//EN
CALSCALE:GREGORIAN
BEGIN:VEVENT
UID:eea4f846-7fd3-4486-a46b-1722803799ca@nexttoyou
DTSTAMP:20261106T173732Z
DTSTART:20261112T183000Z
DTEND:20261112T200000Z
SUMMARY:Conversations on the future of computing: can large language model
 s (LLMs) help us model human choices?
LOCATION:Earthquake Engineering Laboratory (EEL) Auditorium
DESCRIPTION:University of Michigan Civil and Environmental Engineering Cha
 ir Yafeng Yin explores behavioral modeling for civil infrastructure syste
 ms in this presentation.\n\nPeople’s choices shape both their experiences
  with civil infrastructure and system performance\, making behavioral mod
 eling important for design and operation. Yet collecting human choice dat
 a can be costly. Using travel behavior studies\, this talk examines the c
 apabilities and limitations of LLMs as behavioral simulators\, including 
 alignment with human data. Dr. Yin will present two approaches to augment
 ing limited human data for choice model estimation: correcting LLM respon
 ses under uncertainty and transferring utility structure while accounting
  for discrepancies. The central message is that useful LLM augmentation r
 equires explicit discrepancy modeling and validation against human eviden
 ce.\n\nSpeaker bio: Dr. Yin is the Donald Cleveland Collegiate Professor 
 of Engineering and the Donald Malloure Department Chair of Civil and Envi
 ronmental Engineering at the University of Michigan\, Ann Arbor. His rese
 arch examines the interactions among travelers\, infrastructure and mobil
 ity services\, with the goal of improving the design\, operation and regu
 lation of transportation systems. He has published over 160 peer-reviewed
  papers in leading journals and served as editor-in-chief of Transportati
 on Research Part C: Emerging Technologies from 2014 to 2020. He currently
  serves as an area editor for Transportation Science and an as\n\nhttp://
 next-to-u.app/events/eea4f846-7fd3-4486-a46b-1722803799ca
URL:http://next-to-u.app/events/eea4f846-7fd3-4486-a46b-1722803799ca
END:VEVENT
END:VCALENDAR
