Who fuels the fire? Predicting amplifiers, copycats, and attenuators in the networked spread of toxicity in immigration discourse

Drawing on Moral Foundations Theory and recent network diffusion studies, we examine the toxicity spread in US immigration-related discourse on X ( N = 305,878 tweets, January 2021 to April 2024). Using AI-assisted content analysis, this study examines how psychological predispositions (issue stance and moral foundations) and network positions predict users’ toxicity roles (i.e. amplifiers, copycats, and attenuators). Multinomial regression and random forest models reveal that amplifiers are typically centrally embedded users who score high on binding moral values, especially sanctity. Attenuators often hold pro-immigration stances and receive more inbound attention, suggesting reputational restraint and network moderation. Copycats, situated in dense, reciprocated ego networks, tend to reciprocate and mimic the tone of their peers. Contrary to expectations, amplification is not driven by bridge actors but by highly central users. This typology enhances our understanding of how moral convictions and network structures influence propagation, with implications for platform design and network intervention strategies.