Mental models of dynamic systems are different: Adjusting for heterogeneous granularity
本文提出一种新方法，解决心理模型比较中因个体关注细节粒度不同而导致的距离指标夸大问题，通过选择可比变量并压缩链接来保留结构信息，适用于需要分析多人心理模型差异的研究。
This is a methodological contribution to mental model research. It is based on the fact that people emphasize different features of complex situations. Their mental models of the situation are complex because of the situation and of interpersonal diversity. Framed by prior knowledge, they contain elements of distinct detail or granularity levels. Established comparison methods assume that granularity is standardized before elicitation. But unelicited details cannot be analyzed later. However, if elicitation includes details, some of them will be at distinct granularity levels; this leads to unequal distances between some variables. Link-based comparison methods therefore produce exaggerated distance indicators. The method presented here avoids the apparent trade-off between not capturing relevant details and bias from heterogenous granularity. It first selects a subset of variables that are on a comparable level of detail in several mental models, accounting for the frequency of these variables in subgroups. Second, it replaces the sequences of links between each pair of selected variables with a compressed link that maintains the polarity and delay information provided in each mental model. All relevant structural information of the original models is preserved. Such compressed models are constructed for each set of original models to be compared using standard methods without risking to exaggerate distance indicators. Data from a recent study with nine participants illustrates the use.