A105 戦略的相互作用とミクロ・マクロ・ループの自己組織性(マルチエージェント基礎理論,シミュレーションプラットホーム)

概要

論文の詳細を見る
There is no presumption that the self-interested behavior of agents should usually lead to collectively satisfactory results. How well each agent does for it in adapting to its social environment is not the same thing as how satisfactory a social environment they collectively create for themselves. An interesting problem is then under what circumstances will a society of rational agents realize social efficiency? For instance social interactions pose many coordination problems to individuals. They face problems of sharing and distributing limited resources in an efficient way. Consider two parallel routes A and B. Agents that have to travel using one of the route A or B. Agents receive the payoff if they choose the opposite route what the majority does. Solutions to these problems invoke the intervention of an authority who finds the social optimum and imposes the optimal behavior to agents. While such an optimal solution may easy to find, the implementation may be difficult to enforce in practical situations. Self-enforcing solutions, where agents achieve optimal allocation of resources while pursing their self-interests without any explicit agreement with others are of great practical importance. In general, coordination problems are characterized with many equilibria, and then coordination failures result from their independent inductive processes. Many researchers address the problems of how self-interested agents leam to coordinate in a complex and non-stationary worlds. They leave open an important question : what are the underlying strategic interactions that lead to successful coordination or unsuccessful coordination failures? In recent years, this issue has been analyzed as the minority games (MG). However, the growing literature on the MG treats agents as automata, merely responding to changing environments without deliberating about individuals decisions. In this paper we consider the coordination problems of more loosely connected agents, where they behave to satisfy their own goal, and they may coordinate with others in order to improve it. We address the following questions : how do interacting agents manage and self-organize macroscopic orders from bottom up and realize an efficient coordination without any central authority? We specify several learning models both at the individual and the collective levels. Within the scope of our model, we treat models in which agents make deliberate decisions by applying rational procedures, which also guide their reasoning. We explore the mechanism in which interacting agents are stuck at an inefficient equilibrium. While agents understand the that the outcome is inefficient, each acting independently is power less to manage this collective activity about what to do and also how to decide. We propose the evolutionary dynamics in which agents do learn from from the most successful neighbors, and they co-evolve their strategies over time. Each agent adapts the most successful strategy as guides for their own decision (individual learning). Hence their success depends in large part on how well they learn from their neighbors. If the neighbor is doing well, its strategy can be imitated by all others (collective learning).
一般社団法人日本機械学会の論文
2001-11-14

A105 戦略的相互作用とミクロ・マクロ・ループの自己組織性(マルチエージェント基礎理論,シミュレーションプラットホーム)

スポンサーリンク

概要

著者

関連論文

スポンサーリンク