In previous posts we introduced CMAC’s consulting arm, NexusSIM and some of the work it does. This post describes another area of NexusSIM’s work.
NexusSIM offers services in computer simulation, data analysis, and machine learning, with an aim of assisting researchers working toward prosocial outcomes. Most of its work so far has focused on public health research, but it is not limited to that field.
In comparison to typical academic researchers, NexusSIM is distinguished by its emphasis on software engineering tools and standards. While integrating CMAC’s deep expertise in academic research, NexusSIM uses software engineering conventions including project management tools, readability standards in coding work, and unit testing. This makes NexusSIM distinctive in that academic research is often conducted without these protocols, a lack that can ultimately negatively impact the quality of research.
Some of the best examples of NexusSIM’s unique capabilities are in its work with Boston University public health researcher Jonathan Jay. Jay’s research involves testing the efficacy of firearms violence interventions with victims of gun violence who arrive at hospitals. These interventions are typically funded using limited-time government grants, so the longer-term outcomes that such interventions would have in preventing additional firearms violence are difficult to assess.
By using agent-based models, NexusSIM’s computer simulations allow for extrapolating from available data and understanding the impacts of complex demographic and geographic factors on the effects of interventions. This extrapolation enables researchers to account for the probable impacts of interventions in lessening the likelihood that victims of gun violence will be victims or perpetrators of future firearms violence.
In developing a model, NexusSIM was able to adapt an existing model for simulating alcohol-related violence. This involved both changing this model to instead fit the realities of firearms violence and altering the New York City-based assumptions of the model to instead fit census tracts in Boston. In addition to the data from studying firearms violence interventions at hospitals, NexusSIM integrated datasets from the U.S. Census Bureau and from RTI International into the simulations.
This involved attaching variables such as education level and racial identity to simulated agents in specific census tracts. But the model also needed to account for geospatial factors involved in gun violence and its effects on both particular census tracts and other neighborhoods nearby. Additionally, the model integrated factors such as deaths from other causes and accounted for social networks that affect the likelihood of gun violence perpetration or victimization.
By validating the results of agent-based model simulations against existing data, NexusSIM was able to extend the researchers’ capabilities for making accurate claims about intervention effectiveness. This research can be used to target limited funding in ways that will have the best impacts on gun violence prevention.
Contact NexusSIM at info@nexussim.org with any inquiries about how they can assist in research projects that require expertise in computer simulation, data analysis, or machine learning.
