Episodes of sustained protest: temporal patterns of online mobilization on X

The rise of social media has allowed the rapid manifestation of collective actions that are both large in scale and persist over extended periods of time. Yet, little is known about how the mobilization strategies within protests vary over time. By drawing on literature on online mobilization patterns, the concept of cycles of contention and Resource Mobilization Theory, we develop and test a theoretical framework of the development of protest collectives’ communication strategies during episodes of sustained protest. We employ the large language model (LLM) GPT-4 Turbo to analyze a sample of 8,583 posts published on Twitter (now X) by supporters of the 2022 Freedom Convoy in Canada. Our findings show that political mobilization strategies are prevalent at times when there is potential to motivate large numbers of people to join the protest, with coordination efforts being used to facilitate logistics during these times. Conversely, as the protest progresses, shared information and discussions become increasingly marked by anger and fear, which may reflect efforts to sustain participation of a shrinking core of committed supporters. Beyond these theoretical contributions, our study adds to the growing literature demonstrating that LLMs can produce accurate classifications for text as data, even when tasks are complex.