Seminar Abstract

y(1) Introduction to Open Research Center at Kyoto Sangyo University
   and the computerized laboratory for experimental economics.
  (2) Simulating Cheap-talk gamesz

------------------------------
Speaker: Šβθ “Φ

Affiliation:  ‹ž“sŽY‹Ζ‘εŠw ŒoΟŠw•”

Date and Time: 2002.4.18 (Thu), 4:30 - 6:00 p.m.

Place: 3F1136

Chair: HŽR ‰pŽO
------------------------------

ƒ abstract „

(1) The purpose of this project is to advance research in experimental economics 
and the field of economics in general in a laboratory setting. In addition 
this project is concerned with economics education and its impact on society.

(2) This paper describes the behaviour of a simulated player with 
Reinforcement Learning (RL) and comparison between simulations and experiments
n Cheap-talk Games. The simulated player changes her strategy with RL.
We examine how well RL can predict subjects' behaviour, comparing with 
experimental results.
   Previous studies revealed RL could describe subjects' behaviour in some 
of games better than the concept of Nash Equilibrium can. However, they have 
examined few games with communication, such as Signalling Game and Cheap-talk Game.
RL can't describe subjects' behaviour in Cheap-talk Game well when a simulated 
player chooses one of her actions with equal probability at the initial trial, 
but RL describes subjects' behaviour relatively better when she has 
bias (expectation) to her strategies at the initial trial.
   We conclude that the expectation (prejudice) establishes communication 
via Cheap-talk and suggest that comparison between experiments and simulation 
should be effective in simulating games, markets and societies.