Civic Ground Truth in News Recommenders: A Method for Public Value Scoring

TL;DR

This paper addresses aligning news recommendation systems with public interest by introducing civic ground truth, a citizen-generated set of value-based labels for news items. The authors propose a dataset-agnostic pipeline that collects a complete one-month News.com.au corpus, enriches headlines with metadata (prominence, sentiment, surprise), and clusters articles to create a stratified labeling framework evaluated by a nationally representative survey. By linking civic orientations to article features, they build a model for semi-supervised extrapolation of public value scores across a broader corpus, enabling civic-aware re-ranking without relying on proprietary viewership data. The discussion situates civic ground truth as a participatory governance mechanism, while acknowledging limitations such as portability to common datasets and survey biases, and argues for its potential to broaden normative evaluation and support policy goals around media pluralism.

Abstract

Research in news recommendation systems (NRS) continues to explore the best ways to integrate normative goals such as editorial objectives and public service values into existing systems. Prior efforts have incorporated expert input or audience feedback to quantify these values, laying the groundwork for more civic-minded recommender systems. This paper contributes to that trajectory, introducing a method for embedding civic values into NRS through large-scale, structured audience evaluations. The proposed civic ground truth approach aims to generate value-based labels through a nationally representative survey that are generalisable across a wider news corpus, using automated metadata enrichment.