Modeling Sexual Violence within the Homelessness Population through an Agent-Based Approach

Abstract

Sexual violence is an unfortunate crime of opportunity, requiring an offender and a victim. We develop a model that simulates the occurrence of sexually violent events in an urban environment. Here, we take the victims to be homeless individuals since studies find that sexual violence is increasing in homeless communities. We modify an agent-based model for residential burglaries to incorporate non-stationary victims. The residential burglary simulations evolve by having the target house attractiveness increase when nearby residences were burgled, modeling the `broken window effect'. In the case of sexual violence, the victim is non-stationary, yet both offenders and victims visit locations, such as homeless shelters, based on their own perceptions of attractiveness. The attractiveness level of a location is affected by previous events, however, victims and offenders will consider an event differently. We allow for different victim behaviors, such as a risky lifestyle or being at the wrong place at the wrong time. We compare the resulting simulations to the events in San Francisco, CA. The simulated results agree with the expected behavior for the different types of victim behaviors. Moreover, the results show that simulated events are concentrated in regions near homeless shelters. We then tested the model with an increase in the generation of victim agents, observing the outcome for an increase in the number of homeless individuals. Initial increases in the victim generation rate substantially increase the overall amount of crime. However, after a certain level, increasing the number of victims does not increase the crime rates.