socialGeneralization

This repository contains all data and code necessary to reproduce the results of Humans flexibly integrate social information despite interindividual differences in reward (Witt, Toyokawa, Lala, Gaissmaier & Wu, 2024), in which we investigated social learning in correlated reward environments. The code is designed to be run from the top directory.

./simulation

has all code used for simulating

• environmentGenerator.R generates correlated environments

• modelSim.py has simulation, parameter generation and necessary helper (GP,UCB) functions

  • modelSim_allVariants.py has the same, but includes legacy models not used for fitting analysis
  • modelSim_Najar.py has the same for a version of the task that has one agent and one expert making optimal choices
  • simmedModels.py uses modelSim.py to generate datasets for analysis

• evoSim* generates evolutionary simulations

  • evoSimSynthesis.py converts this script's output for further analysis

• SGVis.py generates explanatory plots for the models (e.g. Fig. 2)

and everything related to model fitting and recovery

• modelFit.py has the function to calculate negative log likelihoods of parameter sets given data

• modelRecovery.py is the model recovery script

  • mrecov_parameteradd.py converts this script's output for further analysis

• modelFitting* script for model fitting

  • fitting_synthesis* converts this script's output for further analysis

./analysis

analysis scripts

• behav_measures* generates additional behavioural measures for data (search distance, previous reward etc.); already run on the data

• evoSimAnalysis.py converts

• analysis_evo_e1.R has analysis for evolutionary simulations and Exp. 1

• analysis_e2.R has analysis for Exp. 2

• recovery_analyses.R has model and parameter recovery, as well as bounding logic

• suppAnalysis.R has the supplementary analyses

./environments

home to the environments used in the experiment

./Data

data from experiments, fitting data, evolutionary simulations

./plots

all the plots used in the paper + SI