New AI Aims to Spot Fake News — And Explain Why It's Fake
Researchers built a detector that shows its reasoning, so you can judge for yourself.
Fake-news detectors keep getting more accurate, but many remain difficult to interpret and only weakly connected to established theories of persuasion, credibility, and human judgment. This paper develops a theory-informed computational framework that translates cross-disciplinary theories of fake news into measurable features for automated detection and explanation, using statistical techniques and large language models. To build that foundation, the authors conduct a structured cross-disciplinary review of theories from social sciences, psychology, economics, among other disciplines, revealing how fake news persuades and spreads. Experiments on benchmark datasets show the theory-derived features are predictive and provide interpretable, theory-referenced diagnostic signals. Multi-feature models generally outperform individual features, although gains among the strongest small feature combinations are modest. The work highlights the value of interdisciplinary perspectives in building robust and interpretable fake news detection systems, advancing the foundation for human-centered approaches in combating disinformation.
- The researchers turned psychology research on why people believe lies into concrete signals a computer can measure, like emotional language and appeals to authority.
- Detectors combining several signals beat single-signal ones, but the gains were modest — more signals isn't automatically much better.
- The goal is detectors that show their work, so you can judge the reasoning instead of trusting a mystery score.
Why It Matters
Future feeds could flag false claims and show their reasoning, helping you judge what you read.