Exploratory Model Building Tow ard Agent-Based Econom ics Takashi Iba Masaharu Hirokane Hirokane Kawakami Yoshiyasu Takefuji Heizo Takenaka designers@boxed-economy.org http: / / www.boxed-economy.org/ Post-Office Box 17, Fujisawa Keio-mae Post Office: 4250-1 Endo, Fujisawa, Kanagawa, Japan, zip 252-0816 Keio University, Japan
Sim ulation is …
How Building Reality? Possible World Model and Initial condition simulation FACT result
I ncrem ental Developm ent of the sim ulation m odel From the macroscopic viewpoint, t n e m p l o e l d e o v m e D n l o a Feedback i t t n a e l u m m Application e i r s c n e h I t f o Feedback Prediction or Explanation Application Prediction or Explanation
Exploratory Model Building From the microscopic viewpoint, • Model Building • Model Specification – Specification of initial settings • Parameter Estimation • Model Validation
Phase 1 Phase 1 • Model Building • cognitive science • social psychology • consumer behavior • reinforcement • marketing science learning • management science • neural network • artificial intelligence m eta-algorithm adaptive algorithm m odel Model com ponent com ponent com ponent agent m odel com ponent
Phase 1 Phase 1 There are som e alternative m odels.
Phase 1 Phase 1 • Model Building • communication studies • social network analysis • Sociology • distributed artificial intelligence agent m odel relation m odel com ponent com ponent com posed social m odel
Phase 2 Phase 2 • Model Specification
Phase 3 Phase 3 • Param eter Estim ation I nductive Deductive Param eter Estim ation Param eter Estim ation
Phase 3 Phase 3 Deductive Estim ation ( for Param eter or Model)
Phase 4 Phase 4 • Model Validation • Comparison with observed data observed data • Comparison with other economic theories other economic theories • Comparison with empirical tendencies empirical tendencies
Phase 3 Phase 4 Phase 3 Phase 4 Param eter Estim ation and Model Validation
Total Flow of Exploratory Model Building
Autom atic Model Estim ation System inputs model component simulation user component model designs social model designs automatic model estimation system data set parameter range outputs and increments a=20 b=5.34 data set for x=0.02 validation a=98 C=12000 Permission Z>12345 condition some combinations of model settings
For success w ith social sim ulation
“Neo-I nstrum entalism ” • Instrumentalism + • Extensibility of model • Reality of the assumptions
Open the Black Box ! Boxed Econom y Model = miniature of the entire economy
Recommend
More recommend