Agent-Based Model Limits Sexual Assault on College Campuses

By Lina Sorg

Sexual assault has become an increasingly imperative social justice issue in recent years. “One in five women and one in 71 men will be raped at some point in their lives,” Hannah Callender Highlander of the University of Oregon said. “If you think of that as a mathematician and what those numbers mean, that’s outrageous.” The statistic is even higher on college campuses. “20 to 25 percent of college women and 15 percent of college men are survivors of forced sex during their time in college, while two-thirds of college students experience some form of sexual harassment,” she continued. Rape is also the most underreported crime, with 63 percent of sexual assaults going unreported. 

To combat the disturbing prevalence of sexual assault, the University of Portland—along with many other institutions across the country—has adopted the Green Dot Bystander Intervention, a program that challenges the accepted culture of interpersonal “power-based” violence and encourages students to look out and stand up for one another. It is based on the assumption that a community can collectively and systematically reduce the rate of violence via bystander intervention, and provides individuals with the tools and motivation necessary for effective bystander efforts. During a minisymposium at the 2018 SIAM Conference on Applied Mathematics Education, currently taking place in Portland, Ore., Highlander discussed the work of students in the Department of Mathematics to mathematically examine the Green Dot Bystander Intervention. “We sought to develop a mathematical model of bystander violence prevention to help us understand a particular program we developed on campus several years ago and ultimately make it more effective,” she said.

Highlander completed a four-day intensive training about the intervention program, which drove home the severity of campus assault and personalized the issue. “Hearing the statistics is one thing, but making it personal and realizing how it affects your loved ones is another,” she said. She also spoke highly of the two students who expressed interest in pursuing the project, one of whom was a freshman biology major with a minimal mathematical background but a talent for asking probing, powerful questions. “I really value diversity in all aspects of the word, not because it’s cool but because I’m realizing that is a good way to have a really good research experience,” she said of her choice to mentor the biology student.

The Green Dot Bystander Intervention is centered around the three D’s—direct, distract, and delegate— as means of interacting with a perpetrator or victim, diffusing a worrisome situation, and seeking help for an intervention. The titular “green dot” indicates a behavior, choice, or action that encourages a safe environment and expresses intolerance for the types of sexual-based violence that is increasingly apparent on college campuses. The program groups bystander behavior into two categories (proactive and reactive), trains individuals to recognize potentially-violent “red dot” behaviors, and identifies ways to circumvent obstacles. “The main message is not to dictate what you should do, but saying that we can all do something,” Highlander said.

When creating their model, Highlander and her team had to first decide what tools would be most effective. “We determined that agent-based modeling in the NetLogo platform would best be able to capture the individual behaviors involved in violence bystander prevention,” she said. “In this type of platform, it doesn’t matter what your background is because no programming or math is required, so it levels the playing field.” NetLogo also allows for one-at-a-time sensitivity analysis. The model’s basic components include white dotters, green dotters trained in bystander prevention, and red dotters seeking to do harm — all moving throughout a party environment. The group examined the probability that (i) a green dotter will recognize a troublesome situation and (ii) they’ll act upon noticing, and considered both population-dense and sparse situations. “If you train individuals of influence, when other people see those people intervening they will follow suit,” Highlander said, comparing this behavior to the herd-immunity threshold in infectious disease modeling.

She then displayed an animation demonstrating the three dot types moving throughout the environment. When two move next to each other, they become connected and the probability of something happening increases; if a red dotter is part of the pair, the probability of a violent act occurring increases with time. The group adjusted the model’s parameters to see how the animation—and the incidences of violence and prevention—would change.

While many existing preventative efforts have focused on victim-blaming—identifying what the victim was wearing or how much they were drinking as causes for assault—Highlander found that targeting and reducing those instances was not nearly as effective as increasing the efficacy of trained bystanders to intervene. She and her students chose not to include the effects of alcohol in the model because they did not want to label that as part of the problem. They also consciously chose not to assign gender identities to the dots. Ultimately, Highlander and her students hope that their analysis will optimize the efficiency of the Green Dot Bystander Program. They have begun conversations with the violence prevention coordinator on campus to improve current green-dot training, so it seems that they are well on their way.

Lina Sorg is the associate editor of SIAM News.