Using Collaborative Model Building to Better Understand the Mechanisms of Alcohol-Involved Sexual Violence on College Campuses: A Post-hoc Protocol (Preprint)

BACKGROUND The Collaborative Model Building Project to Understand Sexual Violence (CAMPUS) study seeks to address alcohol-involved sexual violence (AISV) by collaboratively developing an agent-based model (ABM) that can support the decisions of college campuses seeking to address this issue among students. As a first step toward ABM development, we used collaborative model building (CMB), an adaptation of group model building, to co-develop a causal loop diagram depicting key causes and effects of AISV and opportunities for intervention. Our goal of co-creating a CLD that can be translated into an ABM to support intervention decision making differentiates our approach from other participatory systems science studies. OBJECTIVE This paper provides a detailed, replicable post-hoc protocol for using CMB to co-create a CLD of AISV on college campuses that can be translated into an ABM to support intervention decision making. METHODS Our approach consisted of four iterative phases that involved ongoing weekly discussion by the project’s core modeling team (CMT), consisting of researchers with systems science and subject matter expertise. In the first phase, we conducted four CMB sessions with three groups of college campus collaborators to develop one preliminary CLD each. Second, our CMT reviewed each variable and causal connection across the CLDs in consultation with peer reviewed literature to help ensure eventual ABM translation. Third, the CMT combined three CLDs into one, identifying specific loops for review by collaborators. Fourth, we conducted a feedback session with collaborators and created a finalized CLD linked with intervention opportunities. RESULTS From January 2013 to March 2025, we engaged with 39 collaborators in Allegheny County, PA across three groups: 1) college campus practitioners (e.g., student life staff) (N=8), 2) undergraduate students (N=12), and 3) a mix of practitioners and students (N=19). The final output of our process was a complete CLD with 26 loops containing 28 variables showing key AISV mechanisms and intervention opportunities. CONCLUSIONS Given our goal of using the finalized CLD for ABM development, our approach emphasized the selection of potentially measurable and modifiable variables and causal connections . Compared with other participatory systems science approaches, our approach also emphasized intervention identification and prioritization.