Abstract
Both my technical and STS projects focus on how young children can be effectively taught to keep themselves safe in a world where firearms are accessible and the risk of gun-related harm is real. My technical capstone project developed an AI-based firearm safety education system combining a humanoid robot, augmented reality, and large language model-driven dialogue to deliver behavioral instruction to children aged four to seven. My STS research paper examined how gun safety education in U.S. elementary schools evolved before and after the 1999 Columbine High School massacre, pointing out the shift from narrow accident-prevention programs to a layered and politically contested system of safety instruction. The two projects are directly connected, where they share the same subject age group, program materials, and the same question of what it means to teach children about the danger of firearms in a culturally divided environment. Each project continuously shaped how I understood the other.
My capstone project addressed the limits of passive instruction in firearm safety education for young children. Behavioral skills training (BST), which structures learning through instruction, modeling, rehearsal, and corrective feedback, is widely recognized as the gold standard for teaching safety behaviors to children, but its delivery depends on trained human facilitators and controlled settings, making it difficult to scale. To address this, my team developed a three-module system integrating a NAO humanoid robot and augmented reality. The robot module delivered instruction aligned with the four-step safety sequence (Stop, Don't Touch, Run Away, Tell a Trusted Adult) using age-adapted scripts and synchronized gestures. An AR training module enabled gesture-based rehearsal with real-time AI coaching, and an AR testing module assessed skill generalization across three simulated domestic scenarios. Expert evaluation by ten professionals in child development, education, and injury prevention showed strong ratings across all modules, but qualitative feedback identified four key areas for refinement: language accessibility for younger children, trauma-informed interaction protocols, AI dialogue consistency, and fuller implementation of the BST cycle.
My STS paper explored how gun safety education in U.S. elementary schools changed following the Columbine High School massacre, arguing that the event turned a narrow field into a layered, politically fractured system. Before 1999, gun safety education was dominated by accident-prevention programs like the Eddie Eagle GunSafe Program, which reduced children's role to a simple four step avoidance response and treated firearms as neutral objects requiring caution rather than as sources of intentional violence. I noted how the Columbine massacre functioned as a focusing event that fundamentally redefined the problem of school gun violence, shifting attention from accidental encounters to the threat of intentional harm from within student populations. This shift was visible in national political discourse, media framing, and the expansion of school safety infrastructure, including the adoption of lockdown drills, violence-prevention curricula, trauma-informed practices, and formal state safety legislation. However, this change did not replace earlier approaches, it layered over them. The product is a system where children may encounter Eddie Eagle's avoidance message, BST-style behavioral rehearsal, lockdown drills, or social-emotional learning programming, implemented by different institutional actors without shared frameworks or coordination. The paper argues that gun safety education should be understood as a constructed knowledge system whose current form reflects an ongoing negotiation among competing political and cultural interests, with important implications for how future standardization efforts should be approached.
Working on both projects simultaneously revealed things that neither would have reached on its own. The most direct crossover came early, when I began gathering existing curriculum and program materials to inform the design of the technical system and found that standardized, publicly available instructional content for this age group was very limited. This difficulty was a symptom of the broken system my STS paper ended up describing. The lack of coordinated content that I encountered as a design constraint for my technical project became the motivating question of my STS research: why had gun safety education for young children remained so inconsistent, and what forces explained this fragmentation? In the other direction, the STS research deepened the decisions I made in building the technical system. Understanding that the Eddie Eagle program had become contested after Columbine, and that its critics argued it actually normalized rather than just regulated firearms, made me pay more attention to how the robot module framed firearms and what assumptions existed in its dialogue. The two projects shared source material, drawing on the same foundational programs, behavioral training literature, and the same debate about whether passive rule instruction actually changes children's behavior in real encounters. The STS paper's conclusion that gun safety education is a negotiated system rather than a solved one ultimately reinforced the value of the technical project, not as a final solution but as a contribution to an ongoing and unresolved effort to give children tools that actually work, all in an attempt to contribute to the overarching goal to keep kids safe from firearms.
