Internet Filtering
by Derek L. Hansen

Internet blocking software, commonly known as Internet filters, are designed to classify Internet content in order to block or allow access to it. There are a wide range of filters with varying
architectures, levels of effectiveness, and features. Though already widely used in schools and libraries, recent federal legislation (namely the Child Internet Protection Act or simply CIPA) has encouraged the use of filters by requiring their installation in order to receive certain types of federal funding. There was some question as to the constitutionality of CIPA, at least as applied to libraries, but the Supreme Court upheld the law in June of 2003 (United States v. ALA, et al. Supreme Court opinions). As many schools and libraries scramble to comply with CIPA (or decide whether the extra funding is worth their compliance), it has become increasingly important to understand the effectiveness of Internet filters and other criteria that should be considered when selecting a specific fitlering product. I have written a brief article titled "CIPA: Which Filtering Software to Use?" for WebJunction (an excellent online forum for librarians) outlining some important considerations when selecting an Internet filter. For a more comprehensive understanding of the effectiveness of filters and product selection see the following discussion and resources.

Internet Filtering Effectiveness

Due the vast size of the Internet and its constantly changing nature, it is not possible for Internet filters to perfectly classify content. This results in two types of errors. First, under-blocking occurs when content is not blocked that should be restricted. Second, over-blocking occurs when content is blocked that should not have been restricted. There is often a tradeoff between these types of errors (similar to the tradeoff between recall and precision in information retrieval), although specific filtering products can have both lower under-blocking and over-blocking error rates than competing products if their classification is more thorough, timely, and/or accurate.

The effectiveness (lack of over- and under-blocking errors) of Internet filters has been a hotly debated issue and one of great importance as demonstrated by the focus on empirical studies of filtering effectiveness in the ALA case findings of fact of the District Court in 2002 (United Stated v. ALA, et al. District Court opinion - see  section 4. Attempts to Quantify Filtering Programs’ Rates of Over- and Underblocking). Although a number of empirical studies have been performed in the past, most of them have suffered from methodological flaws in sample selection, classification procedures, or implementation of blocking tests. Results have also been interpreted inappropriately, in part because there are two independent measures of over-blocking that are sometimes confused, and likewise for under-blocking. For a comprehensive discussion of methodological issues related to testing the effectiveness of Internet filters see "Calculating Error Rates for Filtering Software" by Paul J. Resnick, Derek Hansen, and Caroline R. Richardson, which will appear in Communications of the ACM (for a preprint version click here).

The study "See No Evil: How Internet Filters Affect the Search for Online Health Information" funded by the Kaiser Family Foundation employs the methodology recommended in the prior article. The study focuses on health information, although a similar methodology can be used to test the effectiveness of other types of information. Also at the Kaiser site, there is an insightful webcast including a presentation of the results and a discussion of their implications. Participants in the webcast include Caroline Richardson, M.D., University of Michigan Medical School; Vicky Rideout, M.A., Kaiser Family Foundation; Emily Sheketoff, American Library Association; Donna Rice Hughes, Enough is Enough; Nancy Willard, Responsible Netizen; David Burt, N2H2 filtering company, and Michael McGee, Planned Parenthood. Two other sources related to this study include a research article (Richardson, et al. 2002, "Does Pornography-Blocking Software Block Access to Health Information on the Internet?" JAMA. 288:2887-2894) and a site including supplemental material (e.g., source data and analysis files) related to the study.

Selecting an Internet Filtering Product

Internet filtering software products have changed dramatically since their inception. There are now a wide variety of products with a host of features, making selection of a particular product difficult. For a brief discussion of important issues to consider when selecting a specific Internet filter see my WebJunction article "
CIPA: Which Filtering Software to Use?" A more thorough synopsis of issues to consider can be found in my presentation slides "Which Filter to Use? Selecting, Configuring, and Evaluating Internet Filters" developed for a PaLA 2003 workshop for librarians (click here to download the entire .ppt presentation 1.5 Mg). I have also created a brief Filtering Software Selection Aid to help compare products and think through important issues when deciding which product to purchase (if any). The ALA is also in the process of finalizing similar tools and checklists which can be found at the ALA CIPA site. Also, see libraryfiltering.com for an excellent, library-focused comparison of specific products.

Contact Information
Derek L. Hansen
University of Michigan
School of Information
email: shakmatt at umich.edu
phone: 734.764.1008
Derek Hansen's CV

Last Updated Feb. 2004