This study sought to develop an AI-integrated verbal violence prevention program for third-grade middle school students and to rigorously evaluate its efficacy. Verbal violence, the most prevalent form of school violence, is characterized by its high ...
This study sought to develop an AI-integrated verbal violence prevention program for third-grade middle school students and to rigorously evaluate its efficacy. Verbal violence, the most prevalent form of school violence, is characterized by its high contextual dependency and manifestation in indirect forms beyond explicit profanity. These complexities have posed significant challenges for traditional, one-way educational interventions, which have shown limited success in fostering meaningful changes in students' awareness and behavior.
To address these limitations, this research introduces a novel pedagogical approach that empowers students to actively recognize and classify verbal violence. The methodology incorporates key data science concepts, including data analysis processes, decision tree modeling, and prompt engineering. Following the PDIE model, a comprehensive six-session program was designed and structured into three distinct phases. The initial phase focused on analyzing verbal violence data to identify underlying patterns. The second phase involved students constructing decision trees and classifying data based on their models. In the final phase, students engaged in prompt engineering to create and utilize their own "clean bots" with ChatGPT.
The program's field applicability and content validity were confirmed through two rounds of review by a panel of five educational experts. Its effectiveness was then assessed with a sample of 81 third-grade students from a middle school in Seoul, who were divided into an experimental group (n=38) and a control group (n=43). To measure changes in potential verbal violence, Jung Da-hye's (2018) scale was utilized, focusing on the sub-factors of impulsive aggression, self-esteem, and depression. Qualitative data were also gathered through in-depth interviews with participants in the experimental group.
The quantitative results indicated that the experimental group exhibited statistically significant reductions in both impulsive aggression (t=2.482, p=0.018) and depression (t=2.757, p=0.009). Conversely, the control group, which received a conventional lecture-based prevention program, showed a significant decrease only in impulsive aggression, with no discernible change in depression levels. While further analysis using ANCOVA and difference-in-differences methods did not reveal statistically significant inter-group differences after controlling for pre-test scores, the qualitative findings confirmed the program's profound educational impact.
Thematic analysis of the interview data revealed that students demonstrated a high degree of engagement with the participatory, AI-driven lessons. They reported experiencing concrete and practical shifts in their perception of verbal violence. Furthermore, the sense of accomplishment derived from constructing decision trees and creating clean bots was suggested as a potential factor contributing to the observed decrease in depression. Some students did, however, mention technical difficulties, highlighting the necessity of foundational instruction in AI-related concepts for the effective implementation of such integrated educational models.
This research holds considerable significance by extending the application of AI-integrated education from its conventional focus on core subjects to the vital domain of safety education, specifically school violence prevention. The study not only confirmed the program's meaningful effects but also demonstrated that students could cultivate AI competencies—such as data analysis, decision tree modeling, and prompt engineering—while simultaneously improving their awareness of verbal violence.
Nevertheless, this study is not without its limitations. The short-term, six-session duration may not have been sufficient to capture more profound, long-term changes in students. Future research should therefore explore extending the application of AI integration to other areas of safety education and, crucially, aim to establish a clearer causal relationship between the sense of achievement experienced in the program and the observed emotional changes. Despite these constraints, this study powerfully illustrates the potential of using AI technology for innovative verbal violence prevention and offers practical implications for the future direction of safety education in the digital era.