HX-Tfd / CSIS

CCS project

Individual Advantage and Group Identities in Agent-Based Opinion Formation

About the project

In this project, we aim to gain a better understanding of opinion spreading in medical communities and its dependency on relations between distinct agents in respective city clusters. We model physicians as agents and based on the adoption dates of a new drug and adjacency matrices obtained from data concerning relationships between physicians inseveral cities train a machine learning model consisting of neural network layers with an algorithm that is physically inspired by a process of diffusion.

The goal of it is to fit four parameters we have identified to be influential in opinion spreading, namely three factors of relationship strength between physicians (friendship, advise and discussion relations) and profit constant obtained from adopting an innovative medication. Upon obtaining fitted parameters on different clusters of physicians, we discuss the applicability of the results inprediction of opinion percentages in other clusters over time concerning adoption of new medication by simulating the algorithm with the obtained parameters on other clusters and analyzing the results.

Dependencies

The code has been tested in python 3.8 with the following dependencies:

tqdm
torch/1.7.0
graph-tool/2.4.3
matplotlib/3.4.3
numpy/1.21.2
pandas/1.3.4
seaborn/0.11.0

For graph-tool, please follow the official installation instructions on https://graph-tool.skewed.de/

Run Random Experiments

We provide the opportunity to experiment with different initialization schemes in our code (see experiment.py) and to visualise the simulation dynamically using the graph-tool module (see visualization.py).

To test the simulation with custom parameters, it is possible to modify the default settings in the two scripts mentioned above. Afterwards, one can adjust the global configurations and simulation settings in main.py and simply run the main script after having installed the dependencies.

The functions used for initialization are

• experiment.py

  • initialise_clusters
  • initialise_driving_forces
  • initialise_agents
  • initialise_opinions
  • change_probs

• visualization.py

  • run_simulation

Data-Driven Experiments

The scripts for the data-driven experiments described in the report are:

experiment_data.py
model_data.py
dataio.py

These are refactored from process.ipynb