Voices on the Wall: Analyzing Intimate Partner Violence Graffiti with Predictive Modeling and Machine Learning

Abstract Purpose Graffiti can communicate much about notions of justice. We were interested to see what U.S. restroom graffiti, directly outside of a domestic violence protective order issuing court, might tell us about the experiences of the women and men seeking protection from intimate partner violence (IPV). Specifically, our research addressed four questions: whether the graffiti thematically connects to its proximity to an IPV protective order court; whether it differs measurably from other restroom graffiti; whether it provides potential for mutual support and connections; and whether it conveys victim realities that foster a sense of community among IPV survivors. Method Restroom graffiti was collected in three different locations. One was a women’s restroom outside of a protective order issuing courtroom, one was the men’s restroom outside the same location and the third was a random women’s restroom. We used a language processing model and predictive machine learning models to analyze the data for similarities and differences. Results The analysis explored four primary research questions offering insight into the themes within the content, uncovering measurable differences across locations, emphasizing graffiti’s potential to facilitate connections and mutual support and demonstrating how this type of graffiti contributes to building meaningful communities. Conclusions This study highlights that graffiti outside IPV protective order issuing courtrooms reflects the emotional experiences of victims, creating thematic connections between their shared realities. It reveals that this graffiti is distinct from other restroom graffiti, suggesting that the location shapes its content. The graffiti also appears to facilitate connections, offering victims an outlet for expression. Additionally, it fosters a sense of community among victims, symbolizing solidarity and support in a public yet personal manner.