News credibility and source assessment¶
Methods and research addressing how to evaluate the reliability, trustworthiness, and factual accuracy of news sources and individual articles. This encompasses source-level assessment (rating entire news outlets by editorial standards, factual accuracy, transparency), article-level evaluation (fact-checking, evidence verification), and crowdsourced trust signals.
Key papers¶
- What Was Written vs. Who Read It: News Media Profiling Using Text Analysis and Social Media Context: predicts outlet credibility via combined analysis of article text, YouTube videos, Twitter/Facebook audience demographics, and Wikipedia content
- Horne et al. (2018) — Sampling the News Producers (NELA2017): Characterizes 92 news sources using 130 content-based features (linguistic style, sentiment, engagement, bias, morality), revealing systematic differences between mainstream outlets, hyper-partisan sources, satire, and misinformation producers.
- Nørregaard, Horne, & Adalı (2019) — NELA-GT-2018: Expands source assessment with 713K articles from 194 sources and multi-dimensional ground truth labels from 8 assessment platforms (NewsGuard, Pew, Wikipedia, OpenSources, Media Bias/Fact Check, AllSides, BuzzFeed, PolitiFact).
- Castillo, Mendoza, & Poblete (2011): Early work on real-time credibility assessment of information on Twitter using features beyond content (user characteristics, information propagation dynamics, network patterns).
Related topics¶
- Fake news detection uses credibility assessment as a downstream task
- Datasets and benchmarks provide labeled sources and articles for evaluating credibility systems
- Media characterization quantifies differences between news sources through features and linguistic analysis
- Bias detection and political polarization overlap with source credibility assessment
Notes¶
Source-level credibility assessment trades per-article nuance for scalability — a reliable outlet may publish occasional false claims, and vice versa. Recent work combines source labels with article content features to handle within-source variation.