Organizations: Electrical Engineering and Computer Science, Syracuse University, Syracuse, NY, USA · Department of Communication, University of California, Santa Barbara, Santa Barbara, CA, USA
Disinformation research has produced increasingly accurate automated fake-news detectors, but many systems remain difficult to interpret and are weakly connected to established theories of persuasion, credibility, and human judgment. In this paper, we develop a theory-informed computational framework that translates cross-disciplinary theories of fake news into measurable features for automated detection and explanation through statistical techniques and large language models. To that end, we conduct a structured cross-disciplinary review of theories from social sciences, psychology, economics, among other disciplines that reveal how fake news persuades and spreads, thereby establishing a broad theoretical foundation for computational modeling. Experiments on benchmark datasets show that 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. Our 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.
Figures & tables
Category
Theory/Concept
Compared to real news, fake news should …
Content
Information Processing Theory of Language Intensity Effects [ 21 ] ; Confidence Heuristi [ 22 ] ; Scarcity Principle [ 23 ]
use more intense, sensational, or hyperbolic language and more certain language.
Emotion Theories [ 24 ] ; Negativity Bias [ 25 ]
contain more emotive language, and especially contain negative emotions.
Expectancy Violation Heuristic [ 26 , 27 ]
contain less clear, organized, and accurate (i.e., error-free) writing.
Narrative Paradigm [ 28 , 29 ]
be less coherent and have internal logical contradictions.
Sourcing Theory [ 29 ]
include less supporting evidence and attributions to external sources (e.g., citations, quotes, links, data).
Illusory Truth Effect [ 30 ]
have more repetitive content.
TABLE I: Theoretical predictions to help distinguish fake from real news.
Representation
Corresponding Theory/Concept (Table I )
Category
Confidence Heuristic
Information Processing Theory of Language Intensity; Confidence Heuristic
Content
Scarcity Principle
Scarcity Principle
Content
Negativity Bias
Emotion Theories; Negativity Bias
Content
Coherence
Expectancy Violation Heuristic
Content
Logical Consistency
Narrative Paradigm
Content
Supporting Evidence
Sourcing Theory
Content
TABLE II: Mapping between theory-informed representations and the social-science theories in Table I .
Dataset
Labels
Train
Test
ReCOVery
Truth
966
120
Fake
487
64
Fake_And_Real_News
Truth
1143
597
Fake
1154
551
GossipCop
Truth
6515
3286
Fake
1653
798
TABLE III: Dataset statistics
Fig. 1: Feature importance ranking on the ReCOVery dataset. Source Credibility emerges as the most dominant predictor.
Fig. 2: Feature correlation heatmap on the ReCOVery dataset. Most pairwise correlations fall within a range around zero, indicating that the theory-informed features capture unique information.
Fig. 3: SHAP plot on the ReCOVery dataset. Source credibility stands out as the primary criterion for differentiating fake news from real news.
Fig. 4: Comparing the importance of theories across datasets.
Fig. 5: Distributions of four representative theory-informed features across real and fake news.
Fig. 6: Pairwise accuracy heatmap illustrating the synergistic effects between theory-informed features.
Faculty of Computer Science & Information Technology, Universiti Malaya, 50603, Kuala Lumpur, Malaysia · LIGM, Univ Gustave Eiffel, CNRS, F-77454 Marne-la-Vallée, France