Local News in the Crosshairs: Audience Perceptions of Algorithmic News Sites

• In Event: Digital News and Democracy

Abstract

Amid five decades of diminishing trust in the American press, recent innovations in artificial intelligence (AI) and algorithmically generated content promise to further disrupt the practices of professional journalists and the relationships between news organizations and their audiences. Though news organizations have for some years used AI tools to more quickly detect newsworthy events and comb through public records, recent and widely hyped advances in large language model-based generative AI tools such as ChatGPT could more drastically upend journalistic routines, contribute to the dissemination of misinformation, and place algorithms in competition with journalists. In this project, we examine one such case study.

In recent years, a network of more than 1,100 algorithmically-generated local news sites under the name Metric Media has launched in all 50 states and the District of Columbia. According to reporting from the New York Times and Columbia University’s Tow Center for Digital Journalism, this network has collected lucrative contracts from political candidates and causes to produce partisan and at times false or misleading content alongside their algorithmically-generated content. Given that local news in the United States earns higher and less polarized trust evaluations, Metric Media has been able to mimic the traditional forms of local news sites to peddle an algorithmic, low-quality, and at times partisan product.

In this project, we ask: (1) How do audiences perceive algorithmically-generated Metric Media sites relative to legitimate local news sites? How does pre-existing trust in media affect these relative evaluations? (2) How does content on Metric Media sites relate to content on reputable news sites in the same market?

We have already collected data on 1,130 Metric Media domains, matched them to their purported coverage region (county or group of counties), and linked this data to county-level characteristics from the U.S. Census and newspaper market characteristics from the Expanding News Desert project.

To answer the first research question, we conduct an original survey in which participants are randomly assigned to view either a Metric Media site covering their state or a matched reputable news sites in their state and asked to evaluate each on trustworthiness, reliability, accuracy, informativeness, interest, and partisan bias. While existing research has examined audience perceptions of algorithmically-generated news content produced by reputable news sites, this study is, to our knowledge, the first to examine perceptions of algorithmically-generated sites.

To answer the second research question, we conduct an original content analysis of geographically matched pairs of Metric Media and reputable news sites, scraping content from each group of sites and comparing the prevalence of different topics, frequency of publication, and partisan bias.

The decline in Americans’ trust in the news media and the collapse of local newspapers threaten to fray the relationship between citizens and their elected officials – leaving the institutional architecture of American democracy vulnerable to corruption, misinformation, and demagoguery. Given the increasing opportunity for information technologies to compete with local journalism and undermine democratic institutions, we believe this project has significant implications for the study of political communication.

Authors

• David Beavers, Harvard University

• Kevin DeLuca, Yale University