• confidence bound
• data-driven analysis
• high-order structures
• online political discussions
• opinion dynamics

In Opinion Dynamics, one of the most exploited approaches for understanding the effects of different biases on public opinions - especially political opinions - is through mathematical models of opinion formation, where parameters incorporate psychological factors (e.g., cognitive biases) affecting individual opinion evolution. In this category are included the so-called bounded confidence models where agents are influenced only by peers having an opinion sufficiently close to theirs. Since different models or parameter values can predict different, even opposite, effects of biases on opinions, there is a crucial need for empirical validation to study and quantify socio-psychological and external drivers of the dynamics. The advent of the Internet, followed by the rise of online social media platforms, allowed the collection of new information regarding human interactions that can be exploited to understand the dynamics of opinion formation and diffusion. This thesis readapts a data-driven time-aware methodology that estimates users’ open-mindedness on pairwise interaction networks to higher-order structures, that better capture group dynamics. The information gathered for the data-driven analysis are collected from Reddit and relate to political debates about United-State politics to have an insight about the interactions between the two political parties, Republican and Democratic. The collected data refers to the time interval from January 2017 to July 2019, the first two years and a half of Donald Trump’s presidency. The task of estimating the users’ open-mindedness distribution is carried out on three main controversial discussion topics regarding gun control legislation, minorities rights and general socio-political issues. The results obtained by modelling interactions, both via pairwise networks and via higher-order networks, are analyzed to understand if the two structures allow to capture different users tendencies and social phenomena.