Rising numbers of mass public shootings have renewed urgency among policymakers and the public to identify effective prevention strategies, as 2025 is on pace to reach a post-pandemic high in such incidents. A number of promising practices, such as behavioral threat assessment and Extreme Risk Protection Orders (also known as red flag laws), have shown potential in preventing these incidents before they occur. However, these practices depend on the person of concern being on the radar of the appropriate authorities. Too often, those best positioned to intervene are either unaware of the person of concern or lack a complete picture of the warning signs due to breakdowns in information sharing.

One of the primary ways mass shootings have been prevented is when someone with information about the person of concern’s warning behaviors and/or communications comes forward, allowing the credibility of the threat to be assessed and managed by those with the necessary training and resources. Yet, individuals with this first-hand knowledge do not always take action. For example, after the 2005 mass shooting at Red Lake High School in Red Lake, Minnesota, investigators determined that nearly two dozen people knew about the perpetrator’s attack plans but did not come forward. Research from the National Threat Assessment Center (NTAC), a branch of the US Secret Service, found that in 23 percent of mass public attacks and 66 percent of targeted school violence incidents, at least one person had observed warning signs and did not report them. Even among cases that were successfully averted, 43 percent of observers still did not take action.

With so many individuals in these perpetrators’ lives observing warning signs of impending violence, how can we better leverage existing systems to strengthen prevention efforts?

A Case Study of the Parkland Mass School Shooting

Our latest study, published in the Journal of Criminal Justice, set out to answer this very question. Using a case study approach, we examined the 2018 mass shooting at Marjory Stoneman Douglas High School (MSD) in Parkland, Florida, an event that has drawn considerable scrutiny for the missed opportunities to prevent it. Specifically, we analyzed the perpetrator’s social network, defined here as the individuals—both close, like family and friends, and more distant, including acquaintances and strangers—who encountered him and observed his concerning behaviors.

Drawing from official records obtained from the investigating law enforcement agency, along with court documents and media reports identified through exhaustive open-source searches, we identified 76 individuals in the perpetrator’s social network.¹ We then examined not only the types of warning signs they observed or reported prior to the attack (for example, threatening communications, discussions of violence, and sharing images of firearms or mass shootings), but also how these individuals were connected to one another.

Our analyses revealed several key findings. First, although 76 individuals were identified in the perpetrator’s social network, the network itself was remarkably sparse, with limited connectedness even among those within the same subgroup, such as peers. This fragmentation likely hindered the sharing of information that could have helped bring together disparate pieces of warning information to form a more complete picture of the threat.

Centralized systems for sharing information, such as school or community tip lines, can gather scattered observations from family, peers, and other community members into one actionable repository… Their effectiveness, however, depends on clear guidance for reporters, education about proper use, and a culture that encourages information sharing.

Despite this overall lack of cohesion, two distinct clusters of network members emerged. The first centered around the perpetrator’s mother, who died several months before the shooting, at a time when his behavior was becoming increasingly erratic and his leakage (verbal threats or references to impending violence) also were escalating. Her death not only removed an important source of oversight but also severed key ties among network members who had been connected through her. The second cluster centered around the school and, more specifically, the administrator leading the threat assessment team. While this individual served as the main point for collecting information about the perpetrator’s threats and concerning behaviors, he lacked the necessary training to assess them effectively and facilitate information sharing among other critical stakeholders, including law enforcement and community mental health providers.

Additionally, nearly all of the individuals (97.4 percent) identified in the perpetrator’s social network had observed concerning behaviors or communications. The largest share of network members (24 individuals) fell within a subgroup of friends, peers, and co-workers. Family members (six individuals) had the highest rate of observing warning signs and the greatest variation in the types observed, followed closely by peers. While half of family members reported concerning behaviors they had observed, reporting among peers was notably lower, at just 34 percent.

Ways to Leverage Perpetrators’ Social Networks for Prevention

Building on our analysis of the MSD perpetrator’s social network, these findings highlight concrete ways to improve prevention efforts. Centralized systems for sharing information, such as school or community tip lines, can gather scattered observations from family, peers, and other community members into one actionable repository. These systems can provide a safe reporting environment and ensure timely follow-up by schools, law enforcement, and community services. Their effectiveness, however, depends on clear guidance for reporters, education about proper use, and a culture that encourages information sharing.

Even when reports were made, the low connectedness of the perpetrator’s network often limited their impact. Information about the perpetrator’s warning communications and concerning behaviors was scattered across social circles and agencies, making it difficult to form a complete picture of the threat. This challenge extended to authorities, as multiple schools, local law enforcement, and state or federal agencies often received fragmented information with no mechanism to synthesize it. Lessons from post-9/11 reforms, such as fusion centers, suggest that similar approaches at the school, county, or state level could improve coordination and intervention.

Our findings also emphasize the need to expand awareness about the types of behaviors that should be reported. Observers were more likely to report direct threats or discussions of guns, while other warning signs, such as idolizing violent role models, using hate speech, or posting cryptic online messages, were less often reported. Educating students, families, and community members about the significance of these broader warning signs can reduce reporting hesitancy by those who observe them and galvanize these individuals to report their concerns, thereby giving threat assessment teams a fuller understanding of potential risks.

Finally, these results more broadly demonstrate the value of social network analysis for prevention. By identifying key individuals in a potential perpetrator’s network and encouraging timely reporting, policymakers and practitioners can mobilize everyday people as crime discouragers who help detect threats before an attack occurs. Programs like See Something, Say Something illustrate the power of bystander awareness, and similar approaches can be applied in schools and communities. Continuing research and evaluation of social networks will be essential for improving reporting, coordinating information, and ultimately reducing the risk of mass public shootings.

ABOUT THE AUTHOR(S)

Emily Greene-Colozzi is an assistant professor in the School of Criminology and Justice Studies at the University of Massachusetts Lowell and a member of the Regional Gun Violence Research Consortium (RGVRC).

Jaclyn Schildkraut is the executive director of the RGVRC at the Rockefeller Institute of Government.

Brent Klein is an assistant professor in the Department of Criminology & Criminal Justice at the University of South Carolina and a member of the RGVRC.

Lucia Arrigo is a security professional and research assistant at the University of Massachusetts Lowell and a special contributor to the RGVRC.

Lauryn Krebs is a doctoral candidate in the School of Criminology and Justice Studies at the University of Massachusetts Lowell and a special contributor to the RGVRC.

---

[1] It bears noting that this is likely an underestimation of the size of his network as people who encountered the perpetrator may not have been interviewed by law enforcement or the media and therefore were not able to be captured in our data collection.