Colloquium20150324-Agent-Based Computational Economics:How the Idea Originated and Where It Is Going


Department of Physics, NCU


Agent-Based Computational Economics:How the Idea Originated and Where It Is Going



陳樹衡教授 (Prof. Shu-Heng Chen)

Department of Economics, NCCU


Date 2015.3.24(Tue)

Time 14:00

Place S4-625





One of the greatest changes in social sciences over the last two decades is their increasing reliance on experimental or simulation approaches to unveil the nature of social complexity. This marvellous march is generally known as the computational social science or the social science from bottom up. In economics, this new field has been recognized as agent-based computational economics (ACE). In this 45-minute talk, we open the subject by giving ACE a historical review and tracing its four origins. From a modelling viewpoint, ACE ‘reinvents’ the bazaar by bringing truly decen-tralized procedures into market analysis, from a single market to the whole econ-omy (macroeconomy). The decentralization process is made of various Homo sapi-ens, from entropy-maximizing agents to autonomous agents. In between the spec-trum, the talk reviews how experimental economics and artificial intelligence have shaped the development of ACE. For the former, the talk discusses how ACE models can be used to analyse the economic consequences of cognitive capacity, personal- ity and cultural inheritance. For the latter, the book covers the various tools used to construct artificial adaptive agents, including reinforcement learning, fuzzy de-cision rules, neural networks, and evolutionary computation. The economic (game) theory of network formation arose in the 1990s. While the development of ACEmod-els is also under the influence of this series of development which is now known as network science, the early ACE models, built from the 1970s to the early 1990s, are network-based. They exist in the form of cellular automata. The talk also surveys this origin from Schelling’s pioneering segregation model all the way down to the modern agent-based model of network formation, and extends them to the related complexity issues, including Herbert Simon’s view of evolving hierarchies. If time is enough, the practical uses of ACE models are also briefly introduced, including the econometric modelling of ACE models, the policy design or business simulation of product innovation using ACE models.

Due to time constraint all technical detailswill be skipped. We only give audience a large and general picture of ACE so that the interested and motivated audience can define their own adventure after the talk. The audience who want to get access to more readings or materials are welcome to send the speaker an email.