Suicide is a leading cause of death in the United States. Global safe reporting guidelines for news reports of suicide intend to mitigate associations of increased suicide incidence and stigma. However, recent research suggests more latent patterns in news beyond the guidelines could still contribute to suicide outcomes such as inhibited help-seeking and isolation. Using the Theory of Spiral of Silence to center isolation, we take a mixed-methods approach to analyze 22,021 articles (2020-2024) and use a zero-shot learning large language model (LLM) classifier to detect suppression of four structural circumstances of suicide: financial/job, legal, school, and access to physical/mental healthcare. We find that circumstance disclosure by news publishers diverges by political leaning, financial (p = 0.016), legal (p < 0.001), and school (p < 0.001); and by regionality, legal (p < 0.001) and health (p < 0.001). We qualify mechanisms of suppression using topic modeling and content sharing networks (CSNs). The spiral of silence lens highlights that left leaning publishers are more likely to disclose systemically or socially collective circumstances. In contrast, right leaning outlets suppress those and instead disclose instances that blame individuals for their experiences. Our work highlights how news reporting can downplay structural factors contributing to suicide. Content Warning: This paper discusses suicide deaths reported in news articles and may be sensitive to readers.

In this study, we conducted an online, between-subjects experiment (N = 2,049) to better understand the impact of warning label sources on information trust and sharing intentions. Across four warners (the social media platform, other social media users, Artificial Intelligence (AI), and fact checkers), we found that all significantly decreased trust in false information relative to control, but warnings from AI were modestly more effective. All warners significantly decreased the sharing intentions of false information, except warnings from other social media users. AI was again the most effective. These results were moderated by prior trust in media and the information itself. Most noteworthy, we found that warning labels from AI were significantly more effective than all other warning labels for participants who reported a low trust in news organizations, while warnings from AI were no more effective than any other warning label for participants who reported a high trust in news organizations.

Pink slime news outlets automatically produce low-quality, often partisan content that is framed as authentic local news. Given that local news is trusted by Americans and is increasingly shutting down due to financial distress, pink slime news outlets have the potential to exploit local information voids. Yet, there are gaps in understanding of pink slime production practices and tactics, particularly over time. Hence, to support future research in this area, we built a dataset of over 7.9M articles from 1093 pink slime sources over 2.5 years. This dataset is publicly-available at https://doi.org/10.7910/DVN/YHWTFC.

Discussions of Violence Against Women (VAW) in publicly accessible forums like online news media can influence the perceptions of people and organizations. Language reinforcing stigma around VAW can result in negative consequences such as unethical representation of survivors and trivialization of the act of violence. In this work, we study the presence of stigmatized framings in news media and how it differs based on media attributes like regionality, political leaning, veracity, and latent communities of news sources. We also investigate the interactions between VAW-based stigma and 14 issue-generic policies used to describe political communications. We found that articles from national, right-leaning, and conspiratorial news sources contain more stigma compared to their counterparts. Furthermore, alignment of articles to the issue-generic policies offers the highest explanation for the presence of stigma in news articles. We discuss implications for institutions to improve safe reporting guidelines on VAW.

In this study, we examine the impact of time on state-of-the-art news veracity classifiers. We show that, as time progresses, classification performance for both unreliable and hyper-partisan news classification slowly degrade. While this degradation does happen, it happens slower than expected, illustrating that hand-crafted, content-based features, such as style of writing, are fairly robust to changes in the news cycle. We show that this small degradation can be mitigated using online learning. Last, we examine the impact of adversarial content manipulation by malicious news producers. Specifically, we test three types of attack based on changes in the input space and data availability. We show that static models are susceptible to content manipulation attacks, but online models can recover from such attacks.

In social networks, individuals and systems work side by side. While individuals make decisions to filter or forward information, systems also prioritize and sort information to manage and assist individual information processing. It has long been argued that system level manipulations can reduce access of individuals to novel information. In this paper, we study how sorting of messages in one's inbox can help or hinder access of diverse information in the network through simulation of cognitively bounded actors. We show that first-in-first-out (FIFO) method of message sorting is ideal in bursty information arrival rates and in networks with lower diameter. Last-in-first-out (LIFO) method of message sorting is ideal for streaming information arrival, but leads to information overload in bursty scenarios by creating too many redundant copies of some of the information in the network. In short, the ideal message sorting method that enhances access to diverse information depends on the network type and information access patterns.