A method for exploring attitude systems by combining Belief Network Analysis and Item Response Theory (ResIN)

Belief network analysis (BNA) is a new class of methods with strong potential to research the organization and development of abstract meaning systems. By mapping the attitude system, this method provides a more profound understanding of often “fuzzy” concepts such as ideologies, worldviews, and norm systems. BNA therefore holds potential implications for a plethora of socially relevant issues. For example, by informing the architecture of extreme belief sets or lines of conflict underlying partisan polarization. Despite the huge potential of this approach, it has some major limitations. Indeed, BNA methods start from the simplistic assumption that opposing groups should be perfectly symmetric in their attitudes (e.g. the more democrats are positive, the more republicans should be negative about each topic). Another important aspect of BNA methods is that they are often grounded on new, instead of well-established theories. This sometimes results in problems of interpretation and reliability of the results.In this article, we introduce a new method by combining BNA with item response theory (IRT). We refer to it as the Response-Item Network (or ResIN) method. This method has the advantage of being grounded in the well-developed psychometrics literature. Furthermore, it allows us to analyze attitudes from different groups without assuming symmetric behavior. This allows us to explore more deeply relationships and differences in the attitude system.Besides validating ResIN using IRT, we also test this method on real data, showing that it produces new insights compared to both classical BNA and IRT. Indeed, we are able to easily distinguish attitudes which belong to the republican and to the democrat side, even in counter-intuitive situations. We furthermore validated the reliability of these results by relying on additional data, such as self-identification measurements.