*************************************************
DIMACS Workshop on Bounded Rationality
January 31 - February 1, 2005
DIMACS Center, Rutgers University, Piscataway, NJ
Organizers:
Lance Fortnow, University of Chicago, [EMAIL PROTECTED]
Richard McLean, Rutgers University, [EMAIL PROTECTED]
Daijiro Okada, Rutgers University, [EMAIL PROTECTED]
Presented under the auspices of the Special Focus on Computation and
the Socio-Economic Sciences.
************************************************
Traditionally, economists and game theorists have assumed that
strategic agents are fully rational in the sense that all players can
completely reason about the consequences of their actions. In the last
few decades, a number of game theorists have argued, in part motivated
by experimental results, that human players do not behave in a way
consistent with theoretical predictions. Questions have been raised
regarding the postulate of full rationality and some have attempted
formalization of partially or boundedly rational players and games
played by such players. This research falls under the rubric of
bounded rationality.
If one takes the view that a process of decision making in economic or
other social situations constitutes a computation in a formal sense of
theoretical computer science, then one is naturally led to some notion
of bounded computational power as a formal expression of bounded
rationality. Two important and complementary questions in this line of
inquiry are: (1) What is the computational power required in order to
play a game in a way consistent with full rationality? (2) If players
are limited in their computational power, how different will
equilibrium outcomes be from the fully rational case? With regard to
the first question, some researchers have examined the computational
complexity of finding best responses in games. As to the second
question, a number of researchers have focused on repeated games
played by various types of computing machines with an emphasis on
their role in facilitating cooperative behavior. In one branch of this
work, bounded rationality is interpreted as bounded recall where a
player's strategic options are limited by constraints that are placed
on memories of past actions. A larger literature models bounded
rationality in terms of finite automata. In particular, the strategies
of players are limited to those that are implementable by finite state
automata. Further work that studies strategies implementable by Turing
machines may be found. Most of the aforementioned work has been
carried out by game theorists and, with the exception of a short burst
of activity in the mid-1990's, there has not been a significant amount
of activity in bounded rationality from the computer science point of
view. This workshop will bring together economists and game theorists
interested in bounded rationality, as well as theoretical computer
scientists with experience in limited computational models. It will
explore previous interactions between computer scientists and
economists concerning this topic. It will then address such issues as:
What are the desiderata of a model of bounded rationality? How do
models of bounded rationality affect conclusions of standard models?
What aspects of human behavior have no compelling model of bounded
rationality to explain them? Are there computational models that
properly estimate the computational power of bounded players while
allowing for an analysis that yields useful results?
**************************************************************
Registration Fees:
(Pre-registration deadline: January 24, 2005)
Please see website for additional registration information.
*********************************************************************
Information on participation, registration, accomodations, and travel
can be found at:
http://dimacs.rutgers.edu/Workshops/Bounded/
**PLEASE BE SURE TO PRE-REGISTER EARLY**
*******************************************************************
---------------------------------------------------------------------
The Cryptography Mailing List
Unsubscribe by sending "unsubscribe cryptography" to [EMAIL PROTECTED]
