BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Emergent Communication through Negotiation - Kris Cao\, NLIP\, Uni
 versity of Cambridge
DTSTART:20180511T110000Z
DTEND:20180511T120000Z
UID:TALK100624@talks.cam.ac.uk
CONTACT:Andrew Caines
DESCRIPTION:Multi-agent reinforcement learning offers a way to study how c
 ommunication could emerge in communities of agents needing to solve specif
 ic problems. In this paper\, we study the emergence of communication in th
 e negotiation environment\, a semi-cooperative model of agent interaction.
  We introduce two communication protocols - one grounded in the semantics 
 of the game\, and one which is a priori ungrounded. We show that self-inte
 rested agents can use the pre-grounded communication channel to negotiate 
 fairly\, but are unable to effectively use the ungrounded\, cheap talk cha
 nnel to do the same. However\, prosocial agents do learn to use cheap talk
  to find an optimal negotiating strategy\, suggesting that cooperation is 
 necessary for language to emerge. We also study communication behaviour in
  a setting where one agent interacts with agents in a community with diffe
 rent levels of prosociality and show how agent identifiability can aid neg
 otiation.
LOCATION:FW26\, Computer Laboratory
END:VEVENT
END:VCALENDAR