Improving Representativeness in Online News Ranking and Recommendation Systems for Local News

Back to 2021 Panel Reports

Written by Sarah Gustavus Lim and Ilia Rodríguez, with contributions from Amara Aguilar, Matthew Ballinger, Russell Contreras, Megan Garvey, Shaun Griswold, Gabriella Schwarz, and Andrea Chong Brás

Published: 3 November 2021

From the executive summary: Throughout 2021, the News Quality Initiative (NewsQ) convened a series of panel discussions to discuss the challenge of improving the representation of cultural diversity in local news feeds. Made up of journalists, journalism educators, and journalism scholars from Los Angeles, California, and Albuquerque, New Mexico, the NewsQ 2021 Representativeness panel ultimately agreed that the multiple dimensions, fluidity, and context-dependent character of representativeness complicate the creation of specific metrics of representativeness that could be used as a measurement in online news ranking and recommendation systems.

According to the panelists, certain dimensions of representativeness, such as producing coverage that reflects the demographic diversity of the local community, are typically  considered essential. However, the creation of a truly representative news feed must consider not only to the racial, ethnic, cultural, or demographic diversity of a place, but also consider how inclusive the coverage was of different neighborhoods, the recommendation of a wider variety of local news outlets, and the reporting of a more diverse mix of topics and voices of interest to people in a community.

During their discussions, panelists identified several structural challenges to achieving a representative local news feed. For example, the adequate volume of representative stories needed to populate feeds may simply not exist. This absence is due to several factors, including the economic challenges affecting local outlets, the ensuing proliferation of news deserts, and an overall lack of diversity in local news ecosystems.

NewsQ panelists agreed that an algorithmically generated newsfeed, even when curated lightly by humans, would not truly illustrate the dimensions of representativeness as identified by panelists. Panelists also recognized that focusing on actively curating available local stories cannot solve the deeper, enduring problems of representativeness that have plagued local news coverage for decades.

Despite the challenges, panelists identified several measures to improve local news feeds. First, producers of news platform products should create systems that can identify actual local news. This requires more dialogue between local newsrooms and technologists. Second, since the lack of representation in the news is a legacy issue in journalism, the panelists saw the problem of representativeness in news feeds as a new opportunity to advance the goal of diversification by bringing together a variety of stakeholders — e.g., journalists, newsroom management, professional organizations, journalism schools, computer scientists, technologists, community groups, and audiences — to address the issue in the current context of news recommendation systems.

Read the report here.