Automated influencers: Unravelling the dynamics of social influence by bots (automated agents) with various strategies in online interactions

Automated accounts—or bots—are increasingly prevalent in online environments, yet their influence on human attitudes remains insufficiently understood. Drawing on Social Identity Theory, this paper presents two pre-registered experimental studies examining how different bot strategies influence individuals' environmental attitudes in interactive online settings. The paper explores the efficacy of specific bot strategies—extreme, escalating, mimicking, and scaffolding—on attitude change compared to a control condition; in small group settings (5 participants) in Study 1 (N = 718), and dyads (Study 2, N = 633). Across both studies, we measured shifts in attitude direction, strength, certainty, identification, and perceived normativity. In addition, Study 1 explores whether participants in group settings could identify bots based on their behaviour. Findings suggest that participants can more easily detect a bot when it exhibits less human-like behaviour (extreme bot). Knowledge of a bot's presence increased psychological reactance, often reinforcing pre-existing attitudes. In dyadic contexts, however, bots were more effective at shifting attitudes, particularly when they used subtle strategies like scaffolding or escalating agreement. Interaction with bots also impacted participants' group identification, suggesting bots can indirectly affect the social meaning of attitudes. Together, these studies highlight the nuanced psychological processes through which bots might influence human attitudes, emphasizing the roles of identity alignment, human-likeness, and perceived group norms. The findings have implications for understanding social influence in algorithmically-mediated spaces (such as social media) and inform ethical considerations for the deployment of bots in public discourse.