Natural language research announcements

["Alexandre Klementiev" <klementi AT uiuc.edu>]

Forwarding a talk announcement from Kiran Lakkaraju.

Alex.

Alex, can you send this out to the AI groups? Prof. Piotr
Gmytrasiewicz is giving a talk on POMDPs that will be of interest to
many in the department. The talk is tomorrow though, so quick
advertisement is necessary!

-Kiran
------------------------------------------------------
Title:  Interactive Partially Observable Markov Decision Processes

Prof. Piotr J. Gmytrasiewicz
University of Illinois - Chicago

3pm, December 6th, 2007 in SC 3403
Abstract:

The talk will extend  the  framework of  partially  observable Markov
decision processes  (POMDPs) to multi-agent  settings by incorporating
the  notion of  agent models  into the  state space.   Agents maintain
beliefs over  physical states  of the environment  and over  models of
other agents, and they use  Bayesian updates to maintain their beliefs
over time. The solutions map belief states to actions. Models of other
agents may include their belief  states and are related to agent types
considered in games of incomplete information.  We express the agents'
autonomy by postulating that their models are not directly manipulable.
IPOMDPs complement a  traditional approach  to  interactive settings
which uses  Nash
equilibria  as a  solution paradigm.   We seek  to avoid  some  of the
drawbacks of  equilibria which  may be non-unique  and do  not capture
off-equilibrium  behaviors.   We  do  so  at the  cost  of  having  to
represent,  process and  continuously  revise models  of other  agents.
Since  the  agent's beliefs  may  be  arbitrarily  nested, the  optimal
solutions  to   decision  making  problems   are  only  asymptotically
computable.