A new degree of freedom for opinion dynamics models: the arbitrariness of scales (2010.04788v1)

Abstract: Opinion dynamics models have been developed to study and predict the evolution of public opinion. Intensive research has been carried out on these models, especially exploring the different rules and topologies, which can be considered two degrees of freedom of these models. In this paper we introduce what can be considered a third degree of freedom. Since it is not possible to directly access someone's opinions without measuring them, we always need to choose a way to transform real world opinions (e.g. being anti-Trump) into numbers. However, the properties of this transformation are usually not discussed in opinion dynamics literature. For example, it would be fundamental to know if this transformation of opinions into numbers should be unique, or if several are possible; and in the latter case, how the choice of the scale would affect the model dynamics. In this article we explore this question by using the knowledge developed in psychometrics. This field has been studying how to transform psychological constructs (such as opinions) into numbers for more than 100 years. We start by introducing this phenomenon by looking at a simple example in opinion dynamics. Then we provide the necessary mathematical background and analyze three opinion dynamics models introduced by Hegselmann and Krause. Finally, we test the models using agent-based simulations both in the case of perfect scales (infinite precision) and in the case of real world scales. Both in the theoretical analysis and in the simulations, we show how the choice of the scale (even in the case of perfect accuracy and precision) can strongly change the model's dynamics. Indeed, by choosing a different scale it is possible to (1) find different numbers of final opinion clusters, (2) change the mean value of the final opinion distribution up to a change of $\pm 100 \%$ and (3) even transform one model into another.