Research & Publications
Analyzes 4.5 million tweets during the 2023 Israel-Hamas War to uncover how coordinated social media campaigns shape crisis information environments. Identifies 11 coordinated groups using multimodal methods — finding that manipulation is fragmented, misleading claims concentrate in just three groups, and no single behavioral signal reliably predicts others. Results inform more targeted content moderation strategies.
* Equal contribution
Presents labeled datasets covering 26 information operation campaigns, combining platform-verified IO posts with over 13M posts by 303k control accounts. Enables researchers to study narratives, network interactions, and engagement strategies of coordinated accounts — and to develop and benchmark IO detection algorithms.
Characterizes coordinated reply attacks in influence operations on Twitter, revealing that primary targets are journalists, news media, and politicians. Proposes two supervised ML classifiers — one to detect targeted tweets (AUC 0.88) and one to identify accounts in coordinated attacks (AUC 0.97) — showing that targeted accounts themselves can serve as sensors for influence operation detection.
A survey of knowledge graph embedding models, covering translation-based and neural network-based approaches that represent entities and relations in low-dimensional vector spaces. Discusses applications leveraging deep learning and social media data, addressing challenges of data sparsity and computational complexity in graph-based knowledge representations.
A critical review of computational propaganda — the use of algorithms, automation, and human curation to distribute misleading information on social media for political manipulation. Covers the evolution from classical propaganda theory to modern tactics including social bots and state-organized troll armies used for astroturfing, and surveys bot detection systems developed to counter these threats.
* Equal contribution
Introduces MEIU22, a multi-platform dataset of social media posts about the 2022 U.S. Midterm Elections spanning Twitter, Facebook, Instagram, Reddit, and 4chan. Covers 1,011 candidates from October 1 to December 25, 2022, with keyword-based post linking and open-source pipeline code to support future multi-platform research.
* Equal contribution
Presents a method to estimate user deletion behavior on social media and investigates online manipulation tactics observed through deletion patterns, revealing coordinated inauthentic activity on Twitter.