The contagion of behaviours, ideas and norms in social networks is not limited to humans: artificial agents —bots, recommendation algorithms, avatars— can amplify it markedly. A study published in Nature examines how the presence of non-human agents in a network alters the dynamics of behavioural spread, and finds that their effect is not neutral: they act as accelerators, speeding up diffusion and raising the adoption threshold the network reaches at equilibrium.

The work addresses an open problem in network science and computational sociology: until now, social contagion models treated all nodes as equivalent, or introduced heterogeneity without distinguishing the artificial origin of agents. The authors build a theoretical framework in which artificial agents are characterised by their capacity for repeated interaction, their lack of fatigue and their continuous availability, properties that humans do not share. According to the study, these differences are enough to qualitatively change the outcome of contagion.

The methodology combines simulations of synthetic networks with observational data from digital platforms. In the models, artificial agents are introduced as a variable fraction of the total population, and the authors measure how the probability that a behaviour becomes sustained changes. The results show that even a small fraction of artificial agents —on the order of a few per cent— can appreciably shift the network's transition point, allowing behaviours that would otherwise die out to take hold.

The implications are relevant for platform design, bot regulation and the understanding of polarisation. The authors note that the amplifying effect does not depend on artificial agents having an intent to persuade: their mere presence and interaction pattern suffice. The study leaves open the question of how to mitigate this effect without restricting legitimate interaction, and proposes as a next step validating the framework with higher temporal-resolution field data.