Exploring the effect of measurement error in ABM – New paper

I am thrilled to announce that our team has recently published a paper in Physica A! Our paper delves into the topic of opinion dynamics models, which have a significant potential for studying current issues such as the diffusion of fake news or vaccine hesitancy. However, to date, these models have had little empirical validation, in part due to difficulties in measuring opinions in ways that directly map to representations in the models.

Our work addresses this issue by analyzing the relationship between opinion measurement errors and model predictions, using the Deffuant model as an example. We discuss three types of measurement errors: random noise, binning, and distortions. While the first two are well-known, the third type is mostly unknown outside of psychometrics. We highlight the nature of each error and how they affect the predictions of the Deffuant model.

Our simulations show that the Deffuant model is robust to binning but not to noise and distortions. In fact, distortions lead to a maximum prediction error of 80%, indicating the importance of accounting for measurement error in opinion dynamics models.

We believe that our research on error propagation in opinion dynamics models will contribute to the expansion and development of this field. By testing the reliability and prediction quality of models before testing them against real-world data, researchers can improve the accuracy and effectiveness of these models.

Full article: https://www.sciencedirect.com/science/article/pii/S0378437122006239