News-Recommender-Favors-Liberal

Here are data and codes for our paper submitted to Science Advances:

• News Recommender Favors Liberals: A Causal Framework for Algorithmic Bias

Prepare the dataset

1. Download the MIND dataset from https://msnews.github.io/.
2. Unzip the files MINDlarge_train.zip and MINDlarge_dev.zip.
3. Put the folders MINDlarge_train and MINDlarge_dev into data/mind.
4. Run chmod 777 prepare_data.sh and ./prepare_data.sh in data/mind.

Run the codes

1. Run chmod 777 run.sh and ./run.sh in code to get the effects used in the paper. For convinience, we have already prepared them in ./code/get_effect/result/mind.
2. All the figures and results presented in our paper and SI can be reproduced by running the jupyter notebooks in ./code/figures. You can cancel comments with plt.savefig() and the corresponding figures will be automatedly saved in ./code/figures/figures.

Thanks for using our code!