Explore papers
Filter by research area, year and venue. Every filter is in the address bar, so you can share a view as a link.
Fact-checking & verificationDatasets & evaluation
The Fact Extraction and VERification (FEVER) Shared TaskThis paper presents results of the first FEVER shared task competition, which challenged 23 teams to classify whether human-written factoid claims could be SUPPORTED or REFUTED using evidence retrieved from Wikipedia. The best system achieved 64.21% FEVER score, demonstrating the difficulty of joint evidence retrieval and natural…
Spread & networksDetection methods
Studying Fake News via Network Analysis: Detection and MitigationThis chapter surveys network-based approaches to fake news detection and mitigation on social media, organizing fake news dissemination along three dimensions (content, social, temporal) and introduces methods exploiting network structure: homogeneous networks (friendship, diffusion, credibility) and heterogeneous networks (knowledge,…
Psychology & interventionsPlatforms, policy & society
Hate Lingo: A Target-based Linguistic Analysis of Hate Speech in Social MediaThis paper presents the first large-scale linguistic and psycholinguistic analysis of hate speech categorized by its target: directed (personal attacks) versus generalized (targeting groups). The authors analyze 28,318 directed and ~2,500 generalized hate tweets, revealing that directed hate is more informal, angry, and explicitly…
Detection methodsDatasets & evaluation
Combating Fake News: A Survey on Identification and Mitigation TechniquesThis comprehensive survey of fake news detection and mitigation organizes existing methods into two broad categories: content-based approaches (linguistic and deep learning methods) and feedback-based approaches (exploiting propagation patterns, user responses, and temporal dynamics). The authors consolidate 23+ publicly available…
Spread & networksDetection methods
Rumor Detection on Twitter with Tree-structured Recursive Neural NetworksProposes two variants of recursive neural networks (RvNN)—bottom-up and top-down—that operate on propagation tree structures of tweets to detect rumors. Rather than treating tweets sequentially, the model learns representations that respect how tweets actually propagate in response threads.
Platforms, policy & societyFact-checking & verification
The Science of Fake NewsThis multidisciplinary review establishes a scientific framework for studying fake news, defining it as fabricated information that mimics news media form but lacks editorial norms. The paper identifies two mitigation strategies—empowering individuals through fact-checking and education, and structural platform interventions—while…
Deepfakes & multimodalDetection methods
EANN: Event Adversarial Neural Networks for Multi-Modal Fake News DetectionThe paper proposes EANN, an adversarial neural network that detects fake news on newly emerged events by learning event-invariant features. An event discriminator forces the multi-modal feature extractor to remove event-specific features while preserving shared patterns across events.
Psychology & interventionsPlatforms, policy & society
Belief in conspiracy theories: Basic principles of an emerging research domainThis special issue introduction proposes four foundational principles for the emerging scientific study of conspiracy belief: conspiracy beliefs are consequential (with real impacts on health and relationships), universal (across cultures and historical periods), emotional (driven by sense-making rather than logic), and social (rooted in…
Bots & coordinated campaignsSpread & networks
Bots increase exposure to negative and inflammatory content in online social systemsSocial bots during the 2017 Catalan independence referendum strategically targeted influential humans to amplify negative and inflammatory content. Operating from network peripheries, bots coordinated sentiment-driven amplification that exacerbated polarization, with approximately one in three referendum-related tweets generated by bots.
Spread & networksBots & coordinated campaigns
Influencers in Polarized Political Networks on TwitterThis paper identifies three types of influencers in polarized political Twitter conversations during Brazil's 2016 Dilma Rousseff impeachment: "opinion leaders" who shape discourse within ideological groups, "informational influencers" (media/institutions) drawn into one camp, and "activists" who amplify like-minded messages. The…
Spread & networksPsychology & interventions
The small, disloyal fake news audience: The role of audience availability in fake news consumptionContrary to claims that fake news has reached a broad audience, this empirical analysis of comScore data shows the fake news audience is small and comprises heavy internet users. Applying the Law of Double Jeopardy from media theory, the authors show that fake news attracts disloyal audiences disproportionately from social platforms.
Spread & networksPlatforms, policy & society
Exposure to opposing views on social media can increase political polarizationThis field experiment tested whether exposure to opposing political views on social media reduces polarization. Twitter users (Republicans and Democrats) were randomly offered money to follow bots retweeting messages from opposing politicians and opinion leaders for one month.
Detection methodsLLMs & generative AI
Separating Facts from Fiction: Linguistic Models to Classify Suspicious and Trusted News Posts on TwitterThis paper presents neural network models combining tweet content, social network interactions, and linguistic features (bias markers, subjectivity cues, moral foundations) to classify Twitter news posts as suspicious (propaganda, hoaxes, clickbait, satire) or verified. The models achieve 95% accuracy on binary classification and 0.7…
Detection methodsSpread & networks
"News you don't believe": Audience perspectives on fake newsUsing focus groups and surveys across the US, UK, Spain, and Finland, Nielsen and Graves find that ordinary people see fake news as a spectrum rather than a binary distinction—encompassing poor journalism, political propaganda, misleading advertising, and fabricated content. Audiences experience the "fake news" problem as one of degree,…
Spread & networksDetection methods
This Just In: Fake News Packs a Lot in Title, Uses Simpler, Repetitive Content in Text Body, More Similar to Satire than Real NewsFake news differs systematically from real news in stylistic features, especially titles. Contrary to the assumption that fake news mimics real news, it is actually more similar to satire: packed titles with fewer stop words and more proper nouns, combined with shorter, repetitive body text with simpler language.
Fact-checking & verificationDatasets & evaluation
ClaimBuster: The first end-to-end fact-checking systemFoundational work introducing ClaimBuster, an end-to-end automated fact-checking system. Introduces a dataset of 23,533 human-annotated U.S. debate statements with check-worthiness labels, establishing baselines for automated claim check-worthiness detection and ranking.
Fact-checking & verificationPsychology & interventions
Claim Ranking in Political DebatesEarly work framing check-worthiness as claim ranking; introduces the notion of "importance weighting" for prioritizing claims for fact-checking in political debates. Foundational for understanding which claims warrant verification effort.
LLMs & generative AI
Attention Is All You NeedThe authors introduce the Transformer, a sequence transduction model based entirely on self-attention mechanisms that dispenses with recurrence and convolution. The model achieves state-of-the-art BLEU scores on machine translation tasks while being significantly more parallelizable and requiring less training time.
Detection methodsDatasets & evaluation
A Survey on Hate Speech Detection using Natural Language ProcessingThis survey systematically reviews automatic hate speech detection methods, covering feature extraction techniques (surface features, embeddings, sentiment analysis, lexical resources, linguistic features, knowledge bases), classification approaches (mostly supervised learning with SVMs and neural networks), annotation practices, and key…
Spread & networksDetection methods
Topological Recurrent Neural Network for Diffusion PredictionWang et al. propose Topo-LSTM, a novel LSTM architecture for modeling dynamic directed acyclic graphs (DAGs) to predict which node will be activated next in information cascades. The key insight is that information diffusion on networks should be modeled explicitly as diffusion topologies—graphs showing how information spreads from…
Detection methodsSpread & networks
Some Like it Hoax: Automated Fake News Detection in Social NetworksHoaxes on Facebook can be detected with 99% accuracy using user interaction patterns (who "liked" posts) rather than content analysis. Two algorithms—logistic regression and harmonic boolean crowdsourcing—achieve high accuracy even with minimal labeled training data, and knowledge transfers across different Facebook communities.
Spread & networksDetection methods
Turing at SemEval-2017 Task 8: Sequential Approach to Rumour Stance Classification with Branch-LSTMThis paper proposes Branch-LSTM, a sequential neural model for rumour stance classification in Twitter conversations. By decomposing conversation threads into linear branches and modeling them with LSTM layers, the approach captures conversational context while assigning stance labels (Support/Deny/Query/Comment) to each tweet.
Bots & coordinated campaignsDetection methods
The Paradigm-Shift of Social Spambots: Evidence, Theories, and Tools for the Arms RaceThis paper provides empirical evidence of a paradigm shift in spambot design on Twitter, where new "social spambots" closely mimic genuine user behavior and successfully evade detection. Neither Twitter's automated systems, human annotators, nor established academic detection tools can effectively identify these accounts; however,…
Detection methodsLLMs & generative AI
CSI: A Hybrid Deep Model for Fake News DetectionCSI is a three-module deep learning model that detects fake news by jointly modeling article text, temporal patterns of user engagement (response), and user behavior patterns (source). The Capture module uses LSTM to learn temporal response signals; the Score module learns user suspiciousness from co-engagement patterns; the Integrate…