自由能与演化博弈:自组织系统基本原理的哲学探究  

The New Challenges for Philosophy of Science:Free Energy and Evolutionary Games

作　　者：刘闯 LIU Chuang

机构地区：[1]复旦大学哲学学院

出　　处：《哲学分析》2023年第5期133-141,199,共10页Philosophical Analysis

摘　　要：关于统一对待动/生物知觉与行动的能动推理理论或预测加工理论,其必然性似乎应该从自由能(最小化)原理推导出来,但实际上并不可能。能为这种必然性提供解释的进路,可能在演化博弈论中找到。论证步骤如下:假设某种动/生物在其演化历史中由于某种原因开始采纳能动推理理论中的贝叶斯预测加工模式作为竞争策略来优化指导其知觉与行动,使之相互补助,那么有可能用演化博弈论来建模解释为什么这样的生物在博弈中会胜过采纳了其他策略的动物,或者等价地说,大多数动物个体采纳了这种策略的种群是不会被持其他策略的个体成功入侵的。Regarding the theory of active inference or predictive processing that unified perception andaction,its necessity seems ought to be derivable from the free energy(minimization)principle,but in factit could not be so.However,it seems possible to find an explanation about its necessity in the evolutionary game theory.The outline of our argument goes like this:I suppose a certain organism begining to assume theBayesian predictive processing model in the active inference theory as its strategy for perception and actionand making them mutually complimentary.Then it is possible to use the evolutionary game theory to constructa model to explain how such a kind of organism can win over organisms using other strategies;or equivalently,in a population where the majority of organism uses the strategy in question,then the population would not besuccessfully invaded by organisms using other strategies.

关 键 词：自由能 最小化 预测加工 演化博弈 预测误差 吃惊度 认知科学 脑神经科学 演化生物学 

分 类 号：B1[哲学宗教—世界哲学]