Why Binary Feedback May Be Enough to Train Better Media-Bias Classifiers

The article evaluates NewsUnfold, a human-in-the-loop news reader that collects user feedback to detect media bias. Findings show that ordinary readers can reliably correct misclassified sentences, improve inter-annotator agreement, and enhance model performance beyond existing expert-labeled datasets. The study highlights dataset flaws, design trade-offs, and future steps for scaling feedback systems across news and social platforms.