Data and code from: Bivalent impact of social networks on overarming: Insights on the alignment between social and individual interests

Fu, Feng 1 ; Herron, Michael1 ; Rockmore, Daniel 1

Published Apr 23, 2026 on Dryad. https://doi.org/10.5061/dryad.41ns1rnvj

Data files

Apr 23, 2026 version files 2.33 MB

gitResults.nb

2.13 MB
mathematica.zip

181.23 KB
README.md

7.54 KB
realnetworksimulations.zip

20.63 KB

Abstract

Drawing on evolutionary game theory, we present a stylized model of gun acquisition for individuals in a society where all have the right to bear arms. In our dynamic model, pairwise confrontations with attendant payoffs take place between individuals who are either armed or unarmed. Individual payoffs depend on the probability of confrontation, the choice to arm, and the choices of others. We show that the likelihood of confrontation affects the optimal societal arming rate and the arming rate that occurs in equilibrium. The latter quickly surpasses the former as the probability of confrontation increases, a phenomenon we call overarming. This reflects a misalignment of individual and societal interests. We further show that spatial structures can exacerbate overarming, especially when individuals perceive a deteriorating social environment. Our modeling results highlight the importance of understanding how fundamental behavioral dynamics and network heterogeneity influence individual decisions to acquire firearms.

Associated Manuscript

Fu F, Herron M, and Rockmore D.
"Bivalent Impact of Social Networks on Overarming: Insights on the Alignment between Social and Individual Interests"
Accepted for publication in Science Advances.

Overview

This submission contains the code and supporting data needed to reproduce the results and figures associated with the manuscript. The materials include:

a Wolfram Mathematica notebook for generating figures,
compiled data files used by the Mathematica notebook,
C++ code for simulations on real social networks, and
anonymized network data used in those simulations.

Software Requirements

Wolfram Mathematica 13.0.0.0 (or a compatible later version) is required to open and run gitResults.nb.
A C++ compiler is required to compile and run the simulation code in realnetworksimulations.
Files with extension .xlsx can be opened using spreadsheet software such as Microsoft Excel or LibreOffice Calc, or imported into Mathematica, Python, R, or MATLAB.
Files with extension .txt are plain-text data files and can be opened with any standard text editor.

File Inventory

1. Mathematica notebook

File: gitResults.nb
Description: Wolfram Mathematica notebook used to reproduce main Figures 1–4 and supplementary Figures S1–S5.

2. Mathematica data archive

Archive: mathematica.zip
Description: Compressed archive containing intermediate and compiled data files used by the Mathematica notebook gitResults.nb to reproduce figures and analyses from the study.
Action required: Unzip mathematica.zip before running gitResults.nb.

After unzipping, the directory mathematica/ contains the following files.

`degree.txt`

Text file containing the degree distribution of a simulated scale-free network.

Column 1: degree — node degree (number of network neighbors); unit: count
Column 2: node_count — number of nodes with the corresponding degree; unit: count
Column 3: node_fraction — fraction of all nodes with the corresponding degree; unitless proportion in the range [0, 1]

This file is used to generate the degree-distribution plot shown in Figure S2A.

`degreeeffect.xlsx`

Excel file containing the relationship between node degree and gun ownership in a simulated scale-free network.

Column 1: degree — node degree (number of network neighbors); unit: count
Column 2: individual_count — number of individuals with the corresponding degree; unit: count
Column 3: armed_fraction — fraction of individuals with that degree who are armed at equilibrium; unitless proportion in the range [0, 1]

This file is used to generate the scatter plot of gun ownership rate versus node degree shown in Figure S2B.

`realnetwork.xlsx`

Excel file comparing equilibrium gun ownership rates in simulated and empirical networks across values of the provocation parameter.

Column 1: provocation_probability — probability of provocation in the model; unitless probability in the range [0, 1]
Column 2: armed_fraction_scale_free — equilibrium gun ownership rate in simulated scale-free networks; unitless proportion in the range [0, 1]
Column 3: armed_fraction_montreal — equilibrium gun ownership rate in the empirical Montreal gang network; unitless proportion in the range [0, 1]

This file is used for the comparison of gun ownership rates in simulated and real networks shown in Figure 3.

`rowT00_b000_it00.txt`

Text file containing a time series of gun ownership in a star-graph simulation.

Column 1: time_step — simulation time step (update step in the model); unit: simulation step
Column 2: armed_fraction — fraction of individuals who are armed at that time step; unitless proportion in the range [0, 1]

This file is used to generate the time-evolution plot for gun ownership on a star graph shown in Figure S4.

`snapshotbeta10.txt`

Text file containing a spatial snapshot of individual arming states on a square lattice.

Column 1: x_index — x-coordinate of a lattice site; unit: lattice index
Column 2: y_index — y-coordinate of a lattice site; unit: lattice index
Column 3: state — arming state or strategy code of the individual at that lattice site; categorical code
- Coding legend: 0 = unarmed, 1 = armed

This file is used to generate the spatial configuration shown in Figure 2B.

`vonNeunambeta10.xlsx`

Excel file comparing simulated and theoretical equilibrium gun ownership rates in a lattice population using a von Neumann neighborhood.

Column 1: provocation_probability_sim — probability of provocation associated with the simulation results; unitless probability in the range [0, 1]
Column 2: armed_fraction_sim — simulated equilibrium gun ownership rate; unitless proportion in the range [0, 1]
Column 3: provocation_probability_theory — probability of provocation associated with the theoretical prediction; unitless probability in the range [0, 1]
Column 4: armed_fraction_theory — theoretically predicted equilibrium gun ownership rate; unitless proportion in the range [0, 1]

This file is used to generate the comparison between simulation and theory shown in Figure 2A.

3. Real-network simulation archive

Archive: realnetworksimulations.zip
Description: Compressed archive containing the C++ simulation code and the anonymized real social network data used in the manuscript.
Action required: This archive must be unzipped before running the simulations or accessing the network data.

After unzipping realnetworksimulations.zip, the archive contains:

Directory: realnetworksimulations/
- Description: C++ code used to simulate gun adoption dynamics on real social networks.
Directory: realnetworksimulations/gangnetworkdata/
- Description: Anonymized and coded real social network data files used in the simulations.

Data Accessibility and Organization

The Mathematica notebook depends on precomputed data files and does not regenerate all results from scratch internally. To reproduce the figures successfully, users should preserve the original relative file structure after unzipping the archives.

Instructions for Use

To reproduce figures

Unzip mathematica.zip.
Open gitResults.nb in Wolfram Mathematica.
Ensure that the files in mathematica/ remain in their original relative location.
Evaluate the notebook to reproduce the main and supplementary figures.

To run the real-network simulations

Unzip realnetworksimulations.zip.
Navigate to the realnetworksimulations/ directory.
Compile the relevant C++ source files using a standard C++ compiler.
Run the simulations as described in the source code and directory contents.

Notes

The .xlsx and .txt files in mathematica/ are supporting data files used to recreate figures and summarized outputs.
The real-network data are anonymized and coded for research reproducibility.
File names reflect simulation settings, parameter choices, or analysis types used in the study.

Contact

For questions regarding the code, data, or reproducibility of the results, please contact the authors of the manuscript.