Clustering and stubbornness regulate the formation of echo chambers in personalised opinion dynamics

Social platforms provide means for users to share opinions and influence each other via online social interactions. The substantial amount of information flowing in such social networks calls for algorithms to filter content to facilitate information processing by the users. Therefore, not only the network structure but also the mechanisms behind these algorithms may affect the information exposed to certain individuals leading to the formation of echo chambers (i.e. opinion bubbles). We study a mechanistic model of opinions on clustered dynamic social networks with sorting algorithms. We find that local social clustering is a key structure to form echo chambers and in combination with community structure can further increase polarisation, particularly with reinforcing algorithms. While reinforcement algorithms often increase the formation of echo chambers in social networks, stubborn individuals may reduce this effect in clustered structures. Furthermore, we identify that when opinions are initially clustered, local clustering and community structure make system-wide polarisation less likely with reinforced algorithm partially because one opinion dominates the dynamics. Our findings contribute to understand the effects of clustering and stubbornness in opinion dynamics regulated by opinion reinforcement filtering.