Skip to content

Latest commit

 

History

History
46 lines (35 loc) · 3.62 KB

rule_data.md

File metadata and controls

46 lines (35 loc) · 3.62 KB

Rule Taxonomy and Timelines

Overview

Rule timelines were reconstructed from the Wayback Machine, scraping at-most-weekly snapshots. These rules are then classified into a hierarchical taxonomy of 3 levels and 17 classes.

Taxonomy

Rule Taxonomy

Rules were classified by a GPT-4o-based classification pipeline. For more details, see our paper on arXiv.

Rule Data Download

Our rules data, spanning from 2018-04-23 to 2024-06-20, is available for download as a .csv file here.

Note that due to the infrequency of Wayback Machine snapshots, there is some uncertainty in both when a rule was created and when a rule was removed. To quantify this, we include the timestamps of the Wayback Machine snapshots on either side of a rule's start and end date. For example, if a snapshot taken on Monday does not include Rule X, but the snapshot taken on Wednesday does include Rule X, we know that Rule X was created sometime between Monday and Wednesay. The same holds for a rule being removed. We include the timestamp of the Monday snapshot (earliest_start) and the Wednesday snapshot (latest_start) as these represent the lower and upper bounds, respectively, of a rule's actual (and unknown) creation date).

The file includes each rule in our dataset as a row, with each row having the following columns:

  • subreddit: the subreddit name
  • earliest_start: the lower bound of when the rule was added
  • latest_start: the upper bound of when the rule was added
  • earliest_end: the lower bound of when the rule was removed
  • latest_end: the lower bound of when the rule was removed
  • Prescriptive: a boolean, expressing if the rule tone is Prescriptive
  • Restrictive: a boolean, expressing if the rule tone is Restrictive
  • Post Content: a boolean, expressing if the rule target is Post Content
  • Post Format: a boolean, expressing if the rule target is Post Format
  • User-Related: a boolean, expressing if the rule target is User-Related
  • Not a Rule: a boolean, expressing if the rule is not able to be classified as a rule
  • Spam, Low Quality, Off-Topic, and Reposts: a boolean, expressing if the rule topic is Spam, Low Quality, Off-Topic, and Reposts
  • Post Tagging & Flairing: a boolean, expressing if the rule topic is Post Tagging & Flairing
  • Peer Engagement: a boolean, expressing if the rule topic is Peer Engagement
  • Links & External Content: a boolean, expressing if the rule topic is Links & External Content
  • Images: a boolean, expressing if the rule topic is Images
  • Commercialization: a boolean, expressing if the rule topic is Commercialization
  • Illegal Content: a boolean, expressing if the rule topic is Illegal Content
  • Divisive Content: a boolean, expressing if the rule topic is Divisive Content
  • Respect for Others: a boolean, expressing if the rule topic is Respect for Others
  • Brigading: a boolean, expressing if the rule topic is Brigading
  • Ban Mentioned: a boolean, expressing if the rule topic is Ban Mentioned
  • Karma/Score Mentioned: a boolean, expressing if the rule topic is Karma/Score Mentioned
  • Rule Text: the full text of the rule

For ease of computing, we recommend loading it using Pandas and creating a MultiIndex. This can be done with pd.read_csv('path_to_downloaded.csv').set_index(['subreddit', 'earliest_start', 'latest_start', 'earliest_end', 'latest_end']).