Whereas the past literature on automatic personality synthesis mostly used human-like cues such as synthetic voice, artificial face/head, and gestures, Lee et al. [13] proposed a personality synthesis model for non-humanoid, information seeking mobile agents, in which the migration behavior of the agents is exploited to exhibit their personality. Although they observed promising results from a subjective evaluation study, the study was conducted over only five arbitrary points in the four-dimensional continuous parameter space of their proposed model. This paper presents the results of our comprehensive evaluation of the model proposed by Lee et al. [13]. In the evaluation, 81 systematically generated conditions were evaluated by 818 Japanese participants from a crowd-sourcing service. The analysis suggests that the model can exhibit at least four different personality types and three out of the four behavior parameters are relevant to the four types. From the analysis, we deduce a triad of rules that govern the relationship between the four parameters and the Big Five personality traits.

过去关于自动人格合成的文献大多使用类似人类的线索，如合成语音、人造面部/头部和手势，Lee 等人。[13] 提出了一种用于非人形信息搜索移动代理的个性合成模型，其中利用代理的迁移行为来展示其个性。尽管他们从主观评估研究中观察到了有希望的结果，但该研究仅在他们提出的模型的四维连续参数空间中的五个任意点上进行。本文介绍了我们对 Lee 等人提出的模型进行综合评估的结果。[13]。在评估中，来自众包服务的 818 名日本参与者对 81 个系统生成的条件进行了评估。分析表明，该模型至少可以表现出四种不同的人格类型，并且四种行为参数中的三种与四种类型相关。从分析中，我们推断出控制四个参数与大五人格特质之间关系的三元组规则。