The effectiveness of moderating harmful online content

This repository contains code and data accompanying the publication "The effectiveness of moderating harmful online content".

Reference

@article{doi:10.1073/pnas.2307360120,
  author = {Philipp J. Schneider and Marian-Andrei Rizoiu},
  title = {The effectiveness of moderating harmful online content},
  journal = {Proceedings of the National Academy of Sciences},
  volume = {120},
  number = {34},
  pages = {e2307360120},
  year = {2023},
  doi = {10.1073/pnas.2307360120}
}

Repository Content:

This repository contains the following code scripts:

• scripts/twitter-data-extraction.ipynb: A notebook used to extract data via the Twitter API. For information on accessing Twitter API for academic research, please refer to the official Twitter documentation on academic research access.
• scripts/run_hawkes_pwl.py: A script that fits the data starting from data/twitter-<topic>-hashtag.csv by utilizing the functions in scripts/fit_hawkes_pwl.py. For further information on fitting Hawkes processes, refer to evently, tick, hawkesbook or other packages.
• scripts/plot-contours.ipynb: A notebook that postprocesses the fits, constructs contour plots, and illustrates the fitted data.
• scripts/plot-social-media-dynamics-deletion.ipynb: A notebook used to plot Fig. 1.
• scripts/dsa_functions.py: Additional functions for reading and analyzing data.

The following data and plots are also available:

• data/twitter-climatescam-hashtag.csv – Contains tweet IDs from Twitter associated with the hashtag #climatescam.
• data/twitter-americafirst-hashtag.csv – Contains tweet IDs from Twitter associated with the hashtag #americafirst or #americansfirst.
  Note: These are dehydrated datasets compiled in compliance with the Twitter Terms of Service. Please refer to further documentation on hydration to obtain the underlying data.
• plots/delete-plot.pdf - Fig. 1 - Social Media Dynamics as Self-Exciting Point Process.
• plots/delta-chi20p-deletion.pdf - Fig. 2 (a) - Reaction time $\Delta$ to achieve harm reduction of $\chi=20$%.
• plots/chi-delta24hour-deletion - Fig. 2 (b) - Harm reduction $\chi$ when content is removed within $\Delta=24$ hours.

Fig. 1 - Social Media Dynamics as Self-Exciting Point Process.

Fig. 2 (a) - Reaction time $\Delta$ to achieve harm reduction of $\chi=20$%.

Fig. 2 (b) - Harm reduction $\chi$ when content is removed within $\Delta=24$ hours.

License:

The dataset and the code in this repository are distributed under the General Public License v3 (GPLv3) license. You can find a copy of the license in the LICENSE file included in this repository. If you have any questions regarding licensing or any other questions, please feel free to contact us at [email protected] .