Balancing Transparency and Accuracy: A Comparative Analysis of Rule-Based and Deep Learning Models in Political Bias Classification
Published in Proceedings of the Second Workshop on Social Influence in Conversations (SICon 2024), 2024
Recommended citation: Martinez, M., Schmer-Galunder, S., Liu, Z., Youm, S., Jayawaeera, C., & Dorr, B. (2024). Balancing Transparency and Accuracy: A Comparative Analysis of Rule-Based and Deep Learning Models in Political Bias Classification. In SICon 2024, pp. 102–115. https://aclanthology.org/2024.sicon-1.7/
The unchecked spread of misinformation and increasing political polarization have raised demand for systems that automatically detect political bias in news. This study compares rule-based and deep learning approaches, applying both to CNN (left-leaning) and FOX (right-leaning) articles. It finds that rule-based models perform more consistently and transparently across data conditions, while deep learning models show higher sensitivity to training data and lower explainability. The work offers insight into designing robust, interpretable political bias classifiers.