A well-educated electorate is generally considered to be normatively good for democracy. Previous studies have shown that much political information comes from soft news sources, leading to concerns about the electorate's ability to participate effectively in democratic processes. However, limitations exist with prior works: the discussion around the role of soft news for broader political engagement does make critical assumptions. Specifically, researchers often implicitly assume distinctive contents as intrinsic delineators of hard and soft news categorization. However, this overlooks the possibility that even within soft news outlets, there exist pockets where political events are being covered as "hard" news, revealing a gap in the literature: what is the content of political news in soft news outlets, and how are they fundamentally different from that of hard news, if any? To address this, however, requires a systematic clarification of hard and soft news, which has proved to be complicated. Our work addresses this gap by leveraging a machine learning framework for identifying hard news within soft news outlets. Collectively, our on-going research makes two contributions to the literature: (1) we are the first to conduct a large-scale annotation task to identify hard news content in soft news outlets. In doing so, (2) we provide a systematic clarification of the information found in hard and soft news, agnostic of style and framing.

PDF