In a case study of the reply ecosystem of an official political party account, compute profile-free behavioral features covering text duplication, character-level entropy, timing regularity, reply latency, and post coverage, complemented by a co-commenting network analysis, and group active accounts with unsupervised density-based clustering.
Abstract
Coordinated and bot-like activity on social media is usually studied with supervised detectors that need rich account data such as profiles, timelines, and follower networks, which is increasingly hard to obtain. We ask what can be established about a single account’s reply ecosystem from its publicly visible posts and replies alone, with no profiles, timelines, or follower data. In a case study of the reply ecosystem of an official political party account (23,953 replies by 1985 accounts, December 2025 to January 2026), we compute profile-free behavioral features covering text duplication, character-level entropy, timing regularity, reply latency, and post coverage, complemented by a co-commenting network analysis, and group active accounts with unsupervised density-based clustering. The clustering, combined with two transparent labeling rules, separates three behavioral tiers: templated amplifiers defined by text reuse, persistent responders with human-like text but extreme volume and coverage, and an organic remainder. The two non-organic tiers comprise 5.4% of accounts, yet produce 53.5% of all comments, a composition that is stable under resampling and threshold sensitivity analysis, with a failure mode that is only conservative, since over-strict settings leave a tier unassigned rather than reshaping it. The platform’s own spam flags, never used as input, rise steadily from organic accounts to templated amplifiers, consistent with the behavioral grouping.
A large-scale empirical analysis of more than 50,000 English-language starter packs and over 600,000 associated users shows that starter packs form a highly interconnected ecosystem with substantial overlap across packs that largely reflects pre-existing communities.
Andrea Failla, V. Freitas, Giulio Rossetti et al.· 0 citations
The increasing influence of social media has amplified the risks associated with automated accounts that spread misinformation, manipulate public opinion and carry out malicious activities. To address this challenge, this study presents an explainable, feature-based approach for detecting social bots on X (formerly Twitter) using user-profile information derived from account metadata and content characteristics. We consolidate and extend existing research by bringing together one of the most comprehensive feature sets explored to date, combining raw attributes, features proposed in the literature, and newly introduced credibility and engagement indicators, together with a previously unexploited profile-personalisation signal. Through a feature engineering and selection process that integrates Mutual Information, Random Forest Importance, and SHAP values, we evaluate the contribution of each feature category and assess its generalisation capacity across three benchmark datasets. Our experiments demonstrate that classical machine learning models enriched with the selected features can match or surpass several state-of-the-art approaches while preserving interpretability. Furthermore, we propose and validate, on the more recent and challenging TwiBot-22 dataset, three categories of features (universal, common, and dataset-specific) that provide a transparent and adaptable basis for generalisable bot detection.
Salvador Lopez-Joya, J. A. Diaz-Garcia, M. Dolores Ruiz et al.· Applied System Innovation· 0 citations
This paper studies social bot detection as dossier-based account auditing with large language models and a graph-structured multi-agent framework. From TwiBot-22, we reconstruct graph data into account-level records combining profile metadata, behavioral statistics, contextual cues, and recent tweets. We compare conventional feature-based baselines, a direct zero-shot Single-LLM auditor, and MAG-Bot, a LangGraph-based multi-agent system. Three findings emerge. First, zero-shot Single-LLM auditing is feasible but has recall-related blind spots, especially on sparse, weakly grounded accounts and coherent role-bound personas. Second, role-constrained multi-agent decomposition substantially improves over Single-LLM: on the 585-account test split, MAG-Bot improves accuracy from 0.5846 to 0.7017, recall from 0.5986 to 0.8289, and F1 from 0.6747 to 0.8028. Third, the gain comes mainly from diagnosis-driven strengthening of the behavioral and contextual specialists, not aggregation tricks or post-hoc debate. Multi-agent LLM auditing therefore derives its main value from role-constrained evidence decomposition and blind-spot correction.
Recommendation systems reorganize culture by linking classification, visibility, and value. This article analyzes Douyin’s #tuwei ecology as a socio-technical system in which platform architecture, creator formatting, audience evaluation, and urban-rural symbolic hierarchies interact through recursive feedback. We formalize this system as four coupled operators—commensuration, sign packaging, metricized valuation, and reception feedback—and derive two observable indices: an item-level transfer-efficiency ratio and a class-level diffusion–appreciation asymmetry index. Using 123,300 public #tuwei videos collected between May and November 2024, multimodal coding of 500 videos, and sentiment analysis of 61,284 comments, we identify a stable diffusion-appreciation asymmetry. Practice-oriented rural videos receive relatively strong approval after encounter, but they diffuse less widely than spectacle-oriented performances optimized for rapid recognition, ironic forwarding, and interaction density. Corpus-level tests (χ2, N = 123,300; Cramér’s V≈0.32) and an outlier-sensitivity analysis confirm the pattern independently of the coded-sample size. To explain this pattern, we propose hyperreal differential order (HDO), a systems concept describing how platform infrastructures repackage embedded cultural practices into feed-compatible sign forms and then differentially amplify them across segmented attention circuits. We position HDO against algorithmic visibility, platformization, and metric power, showing what each leaves underspecified. The study offers a mechanism-based account of how short-video systems reorganize vernacular culture, symbolic power, and interpretive authority without relying on overt exclusion.
Rumor events on social media generate opposing camps whose interaction structure is not captured by spreading models or content detectors. This study describes the camp and bridging structure of three rumor events on Sina Weibo, selected from confirmed cases published by the platform’s rumor-refutation channel. Each event is represented as a multilayer interaction network built from repost, comment, and mention relations. Camps are detected by modularity-based community assignment, and overlap is measured through a fractional membership distribution over communities. Three information-theoretic quantities characterize the structure. In the three events, membership entropy separates committed users from bridging users. Structure-to-stance mutual information measures the alignment between interaction communities and text stance. Cross-layer mutual information measures the consistency of camps across interaction types. A co-opetition matrix of mean edge sentiment describes cooperation within camps and competition between camps and identifies alliance structure. In the three events, membership entropy is bimodal, structure-to-stance mutual information is positive and above a permutation null, and the co-opetition matrix has positive diagonal entries. The three events show three distinct temporal patterns, namely a persistent standoff, a hardening toward a single camp after an official correction, and a reversal with an alliance between two camps. The patterns are recovered under a look-ahead-free temporal scheme. In the three events, bridging users hold higher betweenness centrality than non-bridging users. The results describe cross-camp bridging structure in three rumor cascades and connect the structure to a co-opetition reading of camp relations.
As large language model (LLM) agents become routine participants in public online conversation, understanding how their replies reshape human-to-human interaction structure is a central challenge for social computing and platform governance. Yet most post-deployment evaluations focus on toxicity or engagement volume, leaving thread-level conversational structure—a dimension critical to deliberative quality, social capital formation, and equality of voice—largely unmeasured. We address this gap through a large-scale structural evaluation of @CommentR, a production LLM agent serving millions of users on Weibo, China’s leading microblogging platform. Modeling each thread as a directed reply graph, we match over 216,000 real post-deployment threads on strictly pre-anchor covariates and estimate effects on human-only conversational structure using a doubly robust estimator. Because agent replies often arrive before any human-to-human interaction is observed, we introduce lifecycle-aware estimands that distinguish early-stage formation effects from mature-thread rewiring effects. Under conditional ignorability, agent replies reduce reciprocity, increase branching, and reduce geographic homophily in early-stage threads, while degree-corrected bridging remains unchanged—consistent with a hub-and-spoke shift from dialogic exchange toward one-off commenting around a focal reply. In mature threads, eligibility for the incumbent-rewiring analysis is itself reduced by the agent—sustained incumbent exchange becomes less likely—so we report the mature-thread rewiring estimates as bounded rather than point-identified and treat the formation regime as the one in which our structural evidence is secure. The magnitude of the reshaping depends on how the agent answers: more comprehensive, factual replies produce a larger focal shift. Reply-target analysis confirms that a substantial share of human comments redirect toward the agent, reducing lateral human-to-human exchange. These findings demonstrate that public AI agents can reshape not only what people say but how people talk to each other, with direct implications for platform governance, conversational agent design, and the structural monitoring of AI-mediated public discourse.
Dandan Liu, Lihu Pan, Aznul Qalid Md Sabri et al.· ACM Transactions on Social C...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
