Drifting to more extreme but balanced attitudes: Multidimensional attitudes and selective exposure

We present a model of opinion and attitude dynamics that incorporates multi-dimensionality of a single attitude, i.e. attitudes are composed of evaluations of several features. We include selective exposure that is modeled as a hierarchical version of the bounded confidence model. Individuals adapt their feature evaluations if their attitudes and the respective feature evaluation are sufficiently close. We show that this model can explain a balancing principle, such that features associated with the same object tend to be evaluated similarly. We also show that such individuals can get more extreme despite they might even interact only with agents that have less extreme attitudes.