BEGIN:VCALENDAR
VERSION:2.0
PRODID:ILLC Website
X-WR-TIMEZONE:Europe/Amsterdam
BEGIN:VTIMEZONE
TZID:Europe/Amsterdam
X-LIC-LOCATION:Europe/Amsterdam
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700329T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701025T030000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
UID:/NewsandEvents/Archives/2024/newsitem/15294/9-
 December-2024-Language-Evolution-Learning-Tessa-Ve
 rhoef
DTSTAMP:20241125T145806
SUMMARY:Language Evolution & Learning, Tessa Verho
 ef
ATTENDEE;ROLE=Speaker:Tessa Verhoef
DTSTART;TZID=Europe/Amsterdam:20241209T120000
DTEND;TZID=Europe/Amsterdam:20241209T130000
LOCATION:PC Hoofthuis room 6.05, Spuistraat 134, A
 msterdam or online (Zoom Meeting ID: 878 2270 6729
 )
DESCRIPTION:Human cognition constrains how we comm
 unicate. Our cognitive biases and preferences inte
 ract with the processes that drive language emerge
 nce and change in non-trivial ways. A powerful met
 hod to discern the roles of cognitive biases and p
 rocesses like language learning and use in shaping
  linguistic structure is to build agent-based mode
 ls. Recent advances in computational linguistics a
 nd deep learning sparked a renewed interest in suc
 h simulations, creating the opportunity to model i
 ncreasingly realistic phenomena. These models simu
 late emergent communication, referring to the spon
 taneous development of a communication system thro
 ugh repeated interactions between individual neura
 l network agents. However, a crucial challenge in 
 this line of work is that such artificial learners
  still often behave differently from human learner
 s. Directly inspired by human artificial language 
 learning studies, we proposed a novel framework fo
 r simulating language learning and change, which a
 llows agents to first learn an artificial language
  and then use it to communicate, with the aim of s
 tudying the emergence of specific linguistics prop
 erties. I will present two studies using this fram
 ework to simulate the emergence of a well-known la
 nguage phenomenon: the word-order/case-marking tra
 de-off. I will also share some very recent finding
 s where we test for the presence of a well-known h
 uman cross-modal mapping preference (the bouba-kik
 i effect) in vision-and-language models. Cross-mod
 al associations play an essential role in human la
 nguage understanding, learning, and evolution, but
  our findings reveal that current multimodal langu
 age models do not align well with such human prefe
 rences.
X-ALT-DESC;FMTTYPE=text/html:\n  <p>Human cognitio
 n constrains how we communicate. Our cognitive bia
 ses and preferences interact with the processes th
 at drive language emergence and change in non-triv
 ial ways. A powerful method to discern the roles o
 f cognitive biases and processes like language lea
 rning and use in shaping linguistic structure is t
 o build agent-based models. Recent advances in com
 putational linguistics and deep learning sparked a
  renewed interest in such simulations, creating th
 e opportunity to model increasingly realistic phen
 omena. These models simulate emergent communicatio
 n, referring to the spontaneous development of a c
 ommunication system through repeated interactions 
 between individual neural network agents. However,
  a crucial challenge in this line of work is that 
 such artificial learners still often behave differ
 ently from human learners. Directly inspired by hu
 man artificial language learning studies, we propo
 sed a novel framework for simulating language lear
 ning and change, which allows agents to first lear
 n an artificial language and then use it to commun
 icate, with the aim of studying the emergence of s
 pecific linguistics properties. I will present two
  studies using this framework to simulate the emer
 gence of a well-known language phenomenon: the wor
 d-order/case-marking trade-off. I will also share 
 some very recent findings where we test for the pr
 esence of a well-known human cross-modal mapping p
 reference (the bouba-kiki effect) in vision-and-la
 nguage models. Cross-modal associations play an es
 sential role in human language understanding, lear
 ning, and evolution, but our findings reveal that 
 current multimodal language models do not align we
 ll with such human preferences.</p>\n
URL:https://sites.google.com/view/lela-amsterdam
CONTACT:M. Schouwstra at M.Schouwstra at uva.nl
END:VEVENT
END:VCALENDAR