This repo contains files illustrating the data processing for the Collective Constitutional AI project.
The Jupyter notebook attached (ccai_data_processing.ipynb
) takes
- Pol.is export files (
comments.csv
andparticipants-votes.csv
) - A list of participant IDs to remove (
ids-to-toss.csv
) given spammy inputs (criteria described in the notebook)
And outputs clean_comments.csv
and clean_votes.csv
for turning into the Constitution.
- Group 0 here = group A in the Pol.is report (and group 1 = group B).