Multimodal Coordinated Online Behavior: Trade-offs and Strategies

TL;DR

This work tackles the challenge of detecting coordinated online behavior by embracing multimodality through a multiplex coordination-network framework. It systematically compares monomodal and multimodal operationalizations (MONO, INDE, UNFL, MULTI, INTFL) and analyzes how integration level affects recovery of coordinated communities across two Twitter datasets (UK and IORussia). The authors demonstrate that strongly integrated multimodal approaches (MULTI) tend to preserve most monomodal structures while uncovering additional coordination signals, whereas network flattening (UNFL) often loses important information. The study provides methodological guidance for robust multimodal analyses and highlights practical implications for safeguarding platform integrity, with validated coordination signals in a ground-truth–labeled dataset and insights into selection of community-detection algorithms.

Abstract

Coordinated online behavior, which spans from beneficial collective actions to harmful manipulation such as disinformation campaigns, has become a key focus in digital ecosystem analysis. Traditional methods often rely on monomodal approaches, focusing on single types of interactions like co-retweets or co-hashtags, or consider multiple modalities independently of each other. However, these approaches may overlook the complex dynamics inherent in multimodal coordination. This study compares different ways of operationalizing multimodal coordinated behavior, examining the trade-off between weakly and strongly integrated models and their ability to capture broad versus tightly aligned coordination patterns. By contrasting monomodal, flattened, and multimodal methods, we evaluate the distinct contributions of each modality and the impact of different integration strategies. Our findings show that while not all modalities provide unique insights, multimodal analysis consistently offers a more informative representation of coordinated behavior, preserving structures that monomodal and flattened approaches often lose. This work enhances the ability to detect and analyze coordinated online behavior, offering new perspectives for safeguarding the integrity of digital platforms.

Figures (47)

• Figure 1: Coordinated online behavior is inherently multimodal, as users can coordinate through different types of actions.
• Figure 2: Trade-off between multimodality and constraints loosening in coordinated behavior detection. Increasing the level of multimodality enhances the complexity and richness of the analysis but may also restrict the detection to only the most tightly coordinated actors.
• Figure 3: Different operationalizations of multimodality. The colored shapes are the detected communities, and the dashed boxes show on which network the community detection algorithm is performed.
• Figure 4: Metrics comparison between the layers of the UK and Russia coordination network. The computed metrics are on the actors and edges coverage, and Pearson correlation.
• Figure 5: Normalized Mutual Information (NMI) heatmap comparing the communities detected with the Louvain algorithm, considering only those with size greater than 200 for UK, and 70 for IORussia, across all pairs of co-actions.