Category Co-occurrence and Collateral Blocking in School Web Filtering: Exact Counts for Five CIPA Policies Across 103 Million Domains
Description
Web filters decide by category, and many websites belong to more than one. This dataset describes how 57 web-filtering categories overlap across 103,367,510 domains and what that overlap means for school filtering policies. The co-occurrence files give, for every pair of categories, the number of shared domains, the conditional share in both directions, lift and the Jaccard index. They show, for example, that Adult and Gambling sites rarely carry other labels, while Hate/Discrimination often co-occurs with Political and Illegal Drugs with Health & Medicine, which is why blocking those categories affects more legitimate content. Policy tables translate the overlap into outcomes: for five cumulative policies and three multi-label rules, the domains blocked, the protected-label domains blocked, and the mixed domains left open, broken down by domain class (including .edu and US K-12) and by link-authority band. A cost-ordered curve ranks 28 policy categories by how much protected content each one removes. Counts only, no domain names; a codebook is included. Data source: web filtering for schools (https://www.cipawebfiltering.com/guides/why-schools-need-web-filtering.php).