Collaborative Research: HCC: Small: Socio-linguistic modeling to understand the long-term dynamics of news engagement in online media

The transformation of the news ecosystem from traditional print media to online platforms has fundamentally changed how people engage with current events. This decentralization of media has arguably resulted in greater access to diverse and timely information, but it has also led to growing concerns about unreliable information, polarization, and the role that computer-mediated communication plays in fostering these phenomena.

This project advances the understanding of how people interact with news online and how their behaviors evolve over time. By analyzing how people share, support, or criticize news on social media, this research identifies different stages of news engagement, in terms of the types and tone of news people interact with. Understanding these progression stages and the factors that influence them will provide insights into designing online platforms with healthy news ecosystems, reducing the spread of unreliable information while maintaining information diversity.

The project will integrate findings into university courses and community workshops, ultimately fostering a more informed public.

To meet these goals, this project advances computational models of online news engagement through three main research thrusts. First, it develops models to identify various types of news engagement behaviors and their progression stages, innovating advanced language and user modeling techniques to predict future behavior patterns. Second, it establishes a technical framework for estimating causal relationships between different news engagement behaviors, combining natural language processing with causal inference methods to estimate treatment effects from observational data.

Third, the project tests socio-technical hypotheses regarding strong positions on issues, trust, and information reliability using this framework. The research employs multi-year, publicly available data from online social media platforms, enriched with databases of political news sources. Evaluation methods include machine learning metrics, semi-synthetic experiments, and validation and verification through surveys and focus groups.

This comprehensive approach will produce more accurate predictive models and robust causal estimation methods applicable across various domains to study human behavior from online data.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.