Adaptive Causal Coordination Detection for Social Media: A Memory-Guided Framework with Semi-Supervised Learning

Weng Ding, Yi Han, Mujiangshan Wang · 2026

Read Adaptive Causal Coordination Detection for Social Media: A Memory-Guided Framework with Semi-Supervised Learning by Weng Ding, Yi Han, Mujiangshan Wang (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-00400.txt.

Introduction

Weng Ding, Yi Han, and Mu-Jiang-Shan Wang tackle coordinated fake behavior on social media, where linked accounts act together to mislead. Older detectors lean on surface correlation, fixed settings, and heavy hand labeling. Their 2026 paper presents a three-stage framework, called ACCD, that remembers which detection settings worked well and reuses them in new situations.

First, a cross-mapping technique looks for real causal links between accounts. Next, active learning inside a semi-supervised scheme cuts the labeling burden by asking about the most uncertain cases. A final stage checks results against past detection experience. On real datasets, including the Twitter IRA set, the framework scores well above the strongest baseline. It also cuts hand labeling sharply and runs faster.

Catalog record

CREATORWeng Ding, Yi Han, Mujiangshan Wang
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE35 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-00400

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Adaptive Causal Coordination Detection for Social Media: A Memory-Guided Framework with Semi-Supervised Learning Weng Ding, Yi Han, Mujiangshan Wang Abstract Detecting coordinated inauthentic behavior on social media remains a critical and persistent challenge, as most existing approaches rely on superficial correlation analysis, employ static parameter settings, and demand extensive and labor-intensive manual annotation. To address these limitations systematically, we propose the Adaptive Causal Coordination Detection (ACCD) framework. ACCD adopts a three-stage, progressive architecture that leverages a memory-guided adaptive mechanism to dynamically learn and retain optimal detection configurations for diverse coordination scenarios. Specifically, in the first stage, ACCD introduces an adaptive Convergent Cross Mapping (CCM) technique to deeply identify genuine causal relationships between accounts. The second stage integrates active learning with uncertainty sampling within a semi-supervised classification scheme, significantly reducing the burden of manual labeling. The third stage deploys an automated validation module driven by historical detection experience, enabling self-verification and optimization of the detection outcomes. We conduct a comprehensive evaluation using real-world datasets, including the Twitter IRA dataset, Reddit coordination traces, and several widely-adopted bot detection benchmarks. Experimental results demonstrate that ACCD achieves an F1-score of 87.3% in coordinated attack detection, representing a 15.2% improvement over the strongest existing baseline. Furthermore, the system reduces manual annotation requirements by 68% and achieves a 2.8× speedup in processing through hierarchical clustering optimization. In summary, ACCD provides a more accurate, efficient, and highly automated end-to-end solution for identifying coordinated behavior on social platforms, offering substantial practical value and promising potential for broad application. Keywords: Coordinated attack detection; Causal inference; Social media security; Semi-supervised learning; Convergent cross mapping Figure 1 : Motivation of ACCD. From static, correlation-based detection to an adaptive, memory-guided three-stage framework. 1 Introduction The proliferation of coordinated inauthentic behavior on social media—including disinformation campaigns, artificially amplified narratives, and organized harassment—poses a growing threat to public discourse, election integrity, and platform safety. In response, research in social media security has made considerable strides, yielding increasingly sophisticated models capable of strong performance on critical tasks such as coordinated attack detection, malicious account identification, and information operation analysis. State-of-the-art techniques range from causal methods such as Convergent Cross Mapping (CCM) for identifying influence patterns between accounts, to ensemble classifiers including Random Forests for behavior-based categorization, and automated causal inference frameworks that systematize model selection and effect estimation. Despite these advances, prevailing approaches remain largely reliant on correlation-based heuristics and require extensive manual tuning, limiting their robustness and adaptability in real-world settings. Nevertheless, fundamental challenges persist that constrain the effectiveness, scalability, and generalizability of current detection systems. Key limitations include the frequent inability to distinguish genuine causal influence from spurious correlation; heavy dependence on expert knowledge for parameter selection and threshold configuration; the use of static embedding parameters that cannot adapt to diverse or evolving coordination strategies; prohibitive computational costs when applied at scale; and substantial manual annotation burdens that hinder deployment in time-sensitive scenarios. These shortcomings collectively undermine the operational utility of existing tools, especially as malicious actors continuously refine their tactics to evade detection. Recent efforts have sought to address subsets of these issues, yet notable gaps remain. CCM-based detectors improve causal discovery but suffer from fixed embedding dimensions and scale poorly to large user sets. Social footprint classifiers capture nuanced behavioral cues but depend heavily on manual labeling and domain expertise. Automated causal frameworks provide structured model selection but often lack mechanisms for domain-aware validation and still require expert input for reliable configuration. Thus, a comprehensive solution that integrates adaptive causal modeling, label-efficient learning, and automated validation—without sacrificing scalability or accuracy—remains an open need. To bridge this gap, we propose the Adaptive Causal Coordination Detection (ACCD) framework, a novel three-stage system that dynamically learns optimal detection configurations through a memory-guided adaptive process. ACCD is built on three core principles: (1) employing adaptive convergent cross mapping with parameter selection guided by historical performance to robustly infer causal relationships; (2) integrating active learning within a semi-supervised classification pipeline to drastically reduce annotation effort while preserving accuracy; and (3) deploying an experience-driven validation module that automates model selection and thresholding based on past detection outcomes, eliminating the need for manual expert intervention (see Fig. 1 ). We evaluate ACCD comprehensively across multiple widely recognized benchmarks, including the Twitter IRA dataset (containing 2.9 million tweets from known influence operations), Reddit coordination traces, and the TwiBot-20 bot detection benchmark. Our experiments demonstrate that ACCD consistently outperforms state-of-the-art baselines, achieving substantial gains in detection accuracy, annotation efficiency, and computational performance. The primary contributions of this work are summarized as follows: • Systematic critique and adaptive design. We identify and analyze key limitations in existing coordination detection systems and propose a memory-guided adaptive architecture that explicitly addresses fixed parameterization, scalability bottlenecks, and expert dependence. • Integrated three-stage detection framework. We introduce ACCD, a novel pipeline that unifies adaptive causal discovery, semi-supervised classification with active learning, and automated experience-based validation into a cohesive system designed for high accuracy, low manual overhead, and ease of deployment. • Rigorous empirical validation. We establish a thorough evaluation protocol and demonstrate state-of-the-art results across diverse social media datasets, with observed improvements of 15–20% in F1-score, reductions of up to 70% in manual labeling requirements, and computational speedups of approximately 60% via optimized hierarchical clustering. • Practical impact and generality. Collectively, ACCD offers a scalable, accurate, and highly automated approach to detecting coordinated inauthentic behavior on social platforms. By unifying adaptive causal inference with label-efficient learning and self-validating mechanisms, the framework provides both practical utility for real-world deployment and a structured foundation for future research in adaptive security systems. 2 Related Work The field of coordinated social media attack detection has witnessed significant progress in recent years. Existing work can be broadly categorized into three main directions: causal relationship-based detection methods, behavioral pattern classification approaches, and automated causal inference frameworks. 2.1 Causal Relationship-Based Detection Causal relationship-based approaches have emerged as a fundamental paradigm for coordination detection by leveraging the temporal structure inherent in coordinated activities. The Convergent Cross Mapping (CCM) approach [ 17 ] represents a significant advancement in this direction, leveraging state-space reconstruction based on Takens’ embedding theorem [ 18 ] to infer causal influence between users. Recent work by Levy et al. [ 7 ] applies CCM to social media coordination detection by systematically recording user activity timestamps, vectorizing them into fixed-size time series, and computing influence scores via cross-correlation across increasing library lengths. On the IRA dataset, this approach achieves a precision of 80.0%, a recall of 72.0%, and an AUC of 0.7219 for coordination detection. While this method provides a mathematically grounded framework for detecting coordination, it faces scalability challenges due to its O(N^{2}) computational complexity and limited adaptability to varying coordination patterns across different social media contexts, primarily as a result of fixed embedding parameters. 2.2 Behavioral Pattern Classification Behavioral pattern classification represents a complementary approach that focuses on the distinctive activity characteristics differentiating coordinated accounts from legitimate users. The social footprint classification approach [ 10 , 24 , 29 ] addresses troll identification through systematic categorization of accounts into distinct behavioral types. Building on the taxonomy proposed by Linvill and Warren [ 9 ] , researchers have employed Random Forest classifiers [ 4 ] to process extensive datasets, achieving

Read the complete text