Making Mental Models of Dynamic Systems Comparable

This paper deals with a problem of inter-subject comparisons of mental models of dynamic systems (MMDS): different individuals’ MMDSs have varying degrees of detail. This makes established comparison methods exaggerate the differences between MMDSs. The proposed solution has two phases: (1) selecting the subset of variables that are on a comparable level of detail, taking into account the frequency of variables in subgroups of MMDS, the input and output variables, variables needed to conserve loops and distinctive variables which characterize individual MMDSs and (2) replacing the variables and links on the paths between each pair of selected variables by one link with a composite polarity, delay value and weight. It is shown that this method conserves the paths and loops of the original MMDSs, which is the structural information needed to compare the models. It also reduces the size of the compared MMDs. Data from a recent study with nine MMDSs is used to illustrate the method. It is also suggested that the second phase of this method can be used to automatically generate versions of a model at different degrees of detail/aggregation.