Could Social Media Analytics Help Predict Gun Violence? A Columbia Nursing Review Finds Standardized Definitions Are Key
A growing body of research suggests that social media analytics may help provide early warning signals of gun violence. Yet for progress to continue in this research area, consistent data collection and measurement practices are needed, a Columbia University School of Nursing scoping review recommends. The review,“Mapping social media analytics in firearm injury exposure research: a scoping review,” was published in the Journal of the American Medical Informatics Association (JAMIA) on November 24, 2025.
According to KFF, guns are now the leading cause of death for children and teens in the United States. The health policy organization reports that school shootings have more than doubled over the past two decades, affecting 51 out of every 100,000 children between 2020 to 2024, up from 19 per 100,000 in the early 2000s.
These alarming trends highlight the urgent need for innovative tools to understand and prevent gun violence. With the advent of artificial intelligence (AI), researchers have begun using social media data in new ways, mining public posts for community-level predictors of gun violence.
“Through AI techniques, we can use computer models to extract, analyze, and aggregate large amounts of textual data from social media. The models can generate summaries of topics and identify the overall tone of conversations in real-time,” says Michele Flynch, PhD, a postdoctoral research fellow at Columbia Nursing, and the study’s lead author.
Studies included in the review suggest that social media analytics may help identify the community concerns most closely associated with the risk of gun violence such as increases in gang-related activity, neighborhood displacement, or limited access to mental health services. These insights may support the development of community interventions that address underlying issues contributing to gun violence.
Scoping the landscape
To examine the extent of research on this topic, Flynch and her colleagues reviewed 16 prior studies. Their analysis uncovered substantial variation in how “firearm exposure” and “risk” are defined, measured, and interpreted when using social media data. Only two studies attempted to connect social media signals to public health or law-enforcement data—a critical step for integrated surveillance.
“This variation in definitions represents a missed opportunity for developing coordinated, real-time monitoring systems that integrate social media insights with hospital records, crime data, and other public safety indicators,” the review notes.
Additionally, many studies relied on X (formerly Twitter) as a platform most likely because its text-based posts are easier to analyze. However, Flynch and her colleagues find relying on a single platform overlooks the diversity of online communication.
As a final note, Flynch and her colleagues emphasize the need for broader platform inclusion, consistent metrics, stronger linkages with public health data, and sustained data collection—elements essential for ensuring that social media analytics can meaningfully inform public health responses.