Skip to content

Agent-based model investigating how social interactions can result in self-organized behavioral specialization and social network structure.

Notifications You must be signed in to change notification settings

christokita/socially-modulated-threshold-model

Repository files navigation

Socially Modulated Threshold Model

Agent-based model investigating how social interactions can modify response thresholds and result in self-organized behavioral specialization (i.e, division of labor) and social network structure.

Overview

Scripts of the computational model and subsequent simulation data for:

Tokita CK & Tarnita CE. (2020). Social influence and interaction bias can drive emergent behavioural specialization and modular social networks across systems. Journal of the Royal Society Interface 17: 20190564. http://dx.doi.org/10.1098/rsif.2019.0564

This is the living repository of this code. For the public release/archive of the code used in the paper please see release v1.2.0 of this repository.

Components of this repository

This directory has three main files:

  • scripts: contains all scripts for simulation, analysis, and plotting of the computational model and analytical model.
  • output: contains all subsequent derived data from simulations as well as any graphs produced from analysis of the simulation data.
  • SLURM_scripts: contains all scripts for running R code/simulations on Princeton Della Clusters. I recently reorganized files for this release, so be careful with file names/paths in the scripts here.

Scripts are numbered numerically for organization. The scripts folder has the following structure:

  • util folder contains all custom-made functions used throughout the simualtion. This then, in essence, contains the model itself.
  • 0 installs packages necessary for running the model in parallel on computer cluster. Run this on head node of computer cluster first.
  • 1 non-parallel runs of the model.
  • 2 Parallelized runs of the model.
    • 2b allows processing of data produced by these scripts.
  • 3 Parallelized runs of the model for large parameter sweeps across interaction bias (beta) and social influence (epsilon).
    • 3b can allows processing of data produced by simulations.
    • 3c allows plotting of this processed data.
    • 3_para_sweep folder contains extensive scripts for breaking up parameter sweeps into smaller sweeps for faster simulation on multiple nodes.
  • Analyze_ scripts are used for analyzing various simulation data, organized by focus of the analysis.
    • Analyze_DOL_ scripts focus on analyzing task-performance/behavior data.
    • Analyze_Network_ scripts focus on analyzing interactions patterns, i.e., time-aggregated social networks.
    • Analyze_Stimulus analyzes task-need stimuli over time.
    • Analyze_TaskDistribution analyzes the overall frequency of task performance in a run of the model (usually sweeping across one parameter).
    • Analyze_ThresholdDistribution analyzes the distribution of final thresholds in a run of the model (usually sweeping across one parameter).
    • Analyze_ThreshOverTime analyzes the time evolution of thresholds in a single run of a model (usually sweeping across group size).
  • analytical_calcs folder contains R and Mathematica scripts for the analytical calculations detailed in the supplemental information.
  • long_sims folder contains scripts for running simulations of the model on much longer time scales and analyzing effect of long simulation on model results.
  • supp_analysis folder contains various scripts for checking various details of the model (e.g., changing interaction probability, removing threshold limit)
  • testing_scripts folder contains various "sanity-check" simulations to make sure parallelization does not affect random number generator in model. Also, contains a way to rename files.