Perceptrons Play Repeated Games: New Approach to Bounded Rationality

传统经济理论假定，决策者在吸收和处理信息、进行计算或探索其他行动方案的影响方面的能力是没有限制的。然而，在现实中，经济代理人经常面临太复杂的问题，任何个人都无法以这种方式处理。因此，当信息容量无限的假设放松时，检验传统模型的结果是否稳健是很重要的。本项目的重点是分析一种描述这种能力限制的新方法。具体地说，每个决策者由一个简单的计算机网络代表。网络的总计算能力往往超过单个计算机的能力总和。小型但设计良好的计算机网络可以存储和处理大量信息。复杂性的概念将被形式化，处理复杂性的能力有限的影响将被调查。此外，这种网络还将用于研究适应性学习过程。在这个框架中获得的一些初步结果与现有的关于有限理性的文献中发现的结果截然不同。因此，有必要通过批判性地审查现有的对经济主体有限的信息能力进行建模的工具来进一步开展这项研究。

Traditional economic theory presumes that decision makers have no limits on their ability to absorb and process information, to carry out calculations, or to explore the implications of alternative courses of action. In reality, however, economic agents often face problems that are too complex for any individual to treat in this way. Therefore, it is important to examine whether the results of traditional models are robust when the assumption of unlimited informational capacity is relaxed. The focus of this project is to analyze a new way of describing limits on this capacity. Specifically, each decision maker is represented by a network of simple computers. The aggregate computational power of a network often exceeds the sum of the capabilities of individual machines. Small but well designed computer networks can store and process substantial amounts of information. Notion of complexity will be formalized, and the implications of limited capacity to handle complexity will be investigated. In addition, such networks will be used to study adaptive learning processes. Some of the initial results obtained in this framework are strikingly different from those found in the existing literature on bounded rationality. It is then essential to further this research by critically reviewing the existing tools for modeling the limited informational capability of economic agents.