Bots influence opinion dynamics without direct human-bot interaction: the mediating role of recommender systems

Abstract

Bots’ ability to influence public discourse is difficult to estimate. Recent studies found that hyperpartisan bots are unlikely to influence public opinion because bots often interact with already highly polarized users. However, previous studies focused on direct human-bot interactions (e.g., retweets, at-mentions, and likes). The present study suggests that political bots, zealots, and trolls may indirectly affect people’s views via a platform’s content recommendation system's mediating role, thus influencing opinions without direct human-bot interaction. Using an agent-based opinion dynamics simulation, we isolated the effect of a single bot—representing 1% of nodes in a network—on the opinion of rational Bayesian agents when a simple recommendation system mediates the agents’ content consumption. We compare this experimental condition with an identical baseline condition where such a bot is absent. Across conditions, we use the same random seed and a psychologically realistic Bayesian opinion update rule so that conditions remain identical except for the bot presence. Results show that, even with limited direct interactions, the mere presence of the bot is sufficient to shift the average population’s opinion. Virtually all nodes—not only nodes directly interacting with the bot—shifted towards more extreme opinions. Furthermore, the mere bot’s presence significantly affected the internal representation of the recommender system. Overall, these findings offer a proof of concept that bots and hyperpartisan accounts can influence population opinions not only by directly interacting with humans but also by secondary effects, such as shifting platforms’ recommendation engines’ internal representations. The mediating role of recommender systems creates indirect causal pathways of algorithmic opinion manipulation.

Can a Single Line of Code Change Society? The Systemic Risks of Optimizing Engagement in Recommender Systems on Global Information Flow, Opinion Dynamics and Social Structures

“…An other approach mixing recommender systems and opinion dynamics analyzes how these former can be manipulated by bots to influence social media users (Figure 3, topic 5). Pescetelli et al (2022) have demonstrated that bots can shape public opinion indirectly, even without direct human-bot interaction, thanks to their influence on the very workings of a recommendation system. They examine the influence of these entities on shifting average population opinions and manipulating recommendation algorithms' internal representations.…”

A systematic review of Hyperpartisan News detection: A Comprehensive Framework for Definition, Detection, and Evaluation

“…Although they prove to be efficient, this approach does not scale with the voluminous amount of daily news. To address this issue, alternative automated methods, such as Social Network Analysis, Recommender Systems, Bots, or cross-methodology approaches like Pescetelli et al (2022) can be considered. Hyperpartisan detection may be categorized into three macro-approaches: content, source, and user-based (Pitoura et al, 2017).…”

Reducing Exposure to Harmful Content via Graph Rewiring

“…and societies perceive the world and form their opinions [12,21,36,42,44]. In recent years, platforms have come under increasing scrutiny from researchers and regulators alike due to concerns and evidence that their recommendation algorithms create filter bubbles [6,26,28,45] and fuel radicalization [19,27,39,41,49]. One of the main challenges in this context is dealing with content that is considered harmful [4,7,50].…”