Dynamic opinion strategies outperform static ones in social network influence

A new study by Paolo Tarantino, Fabio Mazza, Carlo Piccardi, and Francesco Pierri investigates how a small set of coordinated actors can manipulate opinions in online social networks. The researchers model the problem using Hegselmann-Krause bounded-confidence dynamics, where agents only adjust opinions toward those within a certain similarity threshold. They test two intervention types on weighted LFR benchmark networks with community structure: static strategies, where stubborn agents keep a fixed extreme opinion, and dynamic strategies, where their opinion gradually evolves from moderate to extreme. Multiple node-selection criteria were compared, including degree, PageRank, betweenness, k-coreness, and s-coreness.

The results show a clear advantage for dynamic strategies. By starting moderate and gradually shifting extreme, these agents exploit bounded-confidence to pull intermediate agents into their orbit, extending influence across communities. Static strategies, in contrast, create early opinion separation and achieve limited reach. Notably, dynamic interventions perform strongly even with simple or random node selection, while static strategies require targeted centrality measures for modest gains. The study has direct implications for designing countermeasures against online manipulation, suggesting that gradual persuasion campaigns pose a greater threat than sudden extreme pushes. This work clarifies the interplay between intervention design and target selection in shaping collective opinions.