Social Bot Detection in Online Social Networks

Social bots are automated or semi-automated accounts that interact, post and replicate human behaviour across online social networks. Their presence spans marketing, political campaigning and the spread of misinformation, with profound implications for public discourse, electoral integrity and public health. Detection strategies have evolved from simple rule-based heuristics to sophisticated machine-learning frameworks that leverage account metadata, network topology and content analysis. Contemporary systems integrate linguistic and structural features, exploiting patterns in posting frequency, interaction networks and textual embeddings. Global efforts now focus on scalable, real-time detection to mitigate coordinated manipulation, misinformation propagation and the erosion of trust in digital communication.

Research from Nature Portfolio

Recent studies have quantified the role of bots in amplifying low-credibility information, revealing that automated accounts disproportionately drive early dissemination of dubious content and target high-influence users to maximise reach. Entropy-based models have uncovered clusters of “bot squads” that follow and retweet designated hubs, exposing coordinated propaganda schemes. Investigations using neutral probe bots demonstrate that platform algorithms exhibit limited systemic political bias, yet partisan users, particularly on the conservative spectrum, inhabit denser networks rich in automated followers and are more exposed to unreliable sources. These insights underscore the need for network-aware interventions and algorithmic transparency to curb the unfolding impact of social bots on information ecosystems.

Social Bot Detection in Online Social Networks publication trend

The graph below shows the total number of articles in social bot detection in online social networks across all publications each year (not limited to Nature Index journals).

Technical terms

Social bot: An automated or semi-automated account designed to mimic human behaviours in online interactions.

Linguistic embedding: A vector representation of words or texts capturing semantic and syntactic relationships for use in machine-learning models.

Structural embedding: A method for encoding the local graph topology of a node’s neighbourhood into a vector space to facilitate pattern recognition in networks.

Recurrent neural network: A class of neural network architectures tailored to sequential data, employing memory units to model dependencies over time.

Low-credibility content: Information originating from sources known to lack editorial standards or rigorous fact-checking processes.

References

1. Twitter Bot Detection Using Neural Networks and Linguistic Embeddings. IEEE Open Journal of the Computer Society (2023).
2. Detecting bots in social-networks using node and structural embeddings. Journal of Big Data (2023).
3. The spread of low-credibility content by social bots. Nature Communications (2018).
4. The role of bot squads in the political propaganda on Twitter. Communications Physics (2020).
5. Neutral bots probe political bias on social media. Nature Communications (2021).
6. DeeProBot: a hybrid deep neural network model for social bot detection based on user profile data. Social Network Analysis and Mining (2022).