The Theoretical and Experimental Study on Strategy Behavior under Mechanism for Public Decision-Making

本研究通过分析被试在不同策略下的策略选择，考察了最大和最小两种策略在诱导被试说出真话中的表现。在实验室条件下，对被试人数和策略进行了限制：2名被试分别使用3、5、8和13种策略。在实验中，每个被试的搭档是一台计算机，它在每一轮中随机选择一个策略。通过对实验数据进行聚类分析和自组织映射（SOM）分析，我们发现，在被试一致选择说真话策略的情况下，有相当一部分被试在策略选择上会出现一些反常行为：受试者在激励机制下显性选择忽略克拉克税的策略，在最大最小机制下，受试者通过计算期望效用来选择获得最大利润的策略，在选择真值的随机机制下，通过调整实验环境或决策环境，可以增加讲述策略。但是，在最大最小机制下，由于利益结构中不存在优势策略，很难提高被试诱导诚实行为的绩效。

In this research, we studied performance of Pivotal and Maxmin mechanisms in inducing truth-telling, by analyzing strategies chosen by subjects under each mechanism. Laboratory experiments on decision-making were conducted with limiting the numbers of subjects and strategies : 2 subjects with 3, 5, 8 and 13 strategies. In the experiments, the partner of each subject was a computer which chose a strategy randomly in each round. This experimental method was effective in reducing randomness in strategy-choosing by the side of subject.By applying experimental data sets to cluster analysis and self-organizing maps (SOM), we found that quite a few subjects adopt some anomalous behavior in choosing strategies where subjects are expected to choose unanimously the truth-telling strategy: a subject dominantly chooses a strategy ignoring Clarke tax under the Pivotal mechanism, and a subject chooses a strategy deriving max profit by calculating the expected utility under the Maxmin mechanism.Under the Pivotal mechanism choosing truth-telling strategies may increase by adjusting experimental or decision-making environment. However, it is difficult to improve the performance of subjects in inducing truth-telling behavior under the Maxmin mechanism, because there is no dominant strategy in profit structure in the Maxmin mechanism.