Markkula Center of Applied Ethics

The Effectiveness of Filtering Software

This question has generated the greatest debate, as one can see from scanning the articles in Appendix 14: librarian vs. librarian; government official vs. government official; citizen vs. citizen. Because of this, we made every effort to understand the current state of filtering software.

As one also notices in Appendix 14, filtering software is in its infancy. SafeSurf was the first product and it is just about three years old. With other names like "NetNanny," "CyberSitter," "Surfwatch," and "X-Stop," these products have marketed themselves as convenient ways to keep pornography, pedophiles, and other objectionables away from children. Such products also depict themselves as anti-censorship, and, in general they do not think of themselves as censorship products.

Common descriptions of these products include:

  • "NetNanny's screening lists are completely user defined...according to their particular values-not a developer's arbitrary selection or the Government's!" (NetNanny)

  • "CyberPatrol provides parents, teachers, day care professionals-anyone who is responsible for children's access to the Internet-with the tools they will need to get a handle on an area which can be very dangerous for kids." (CyberPatrol)

  • "SurfWatch Internet filtering products help you with the flood of inappropriate material on the Internet. An alternative to government censorship, SurfWatch provides a technical solution to the difficult issues created by the explosion of technology." (SurfWatch)

Although various surveys and reviews of the different blocking systems have been done or are in process (e.g., the bluehighways project), the results often contradict one another. What is evident from every review is that all Internet blocking mechanisms-even the very best-succeed some of the time but also fail to block sites accurately at least some of the time.

The earlier blocking and filtering software fell (and some continue to fall) loosely into two groups: (a) services that blocked sites containing a word or words considered obscene or evidencing sexually explicit or otherwise objectionable content, and (b) services that had persons exploring and blocking sites individually. Although the former allows users to access far more sites than the latter (as it does not maintain a list of sites, but searches all sites for the "improper" words and word strings), it rarely works very well. In one popular system, all sites containing the word "breast" were blocked, including those dealing with breast cancer (this has since been corrected). In a more recent experiment with SurfWatch, often reputed to be one of the best blocking software, one was able to view graphically-explicit sexual fetish sites, while a New York Times article on Internet gambling was blocked. Both blocking errors are presumably the result of word or word-string searches.

The second type employs actual persons to sort and review sites, but can only get to a fraction of the sites on the Web. Some services, such as the aforementioned SurfWatch and KinderGuard, employ persons to review and block web sites individually (b), using words and word-strings to filter sites they have not reviewed (a). This combination of the two original blocking methods appears to be superior to either method alone. It does not resolve the problems with the two methods; it merely reduces the likelihood of access to objectionable sites while maintaining something of the boundlessness of the Internet. Many blocking software producers are willing to acknowledge this much. Jay Friendland, co founder of SurfWatch, admits, "It's part of a solution. It's not the complete solution."

All the companies we interviewed either have or will soon have the following capabilities: 1.) blocking software with filters as determined by the company; 2.) different levels of blocking, customizable by the user or the network administrator; individual sites can be blocked or unblocked; 3.) software that comes with no sites blocked and network administrators can block sites as they wish; 4.) the availability of third-party filters (so, for example, the ALA or KIDS could develop their own filter files.)

We have included in Appendix 14 the selection criteria we could find for each of the company's filter files. As far as we could tell, all companies use some version of a computer program or an intelligent agent or a "bot" to surf the Web looking for sites that meet specific search criteria we could not discover because they are considered "trade secrets" in a highly competitive industry. For instance, some companies identify the major adult service providers and block the server addresses these companies use (raising the question of whether other, perhaps legitimate users of that server are also blocked); others block all sites that require you to say you are 18 years of age or older; others use pattern matching (certain word phrases) to determine the sites. Even companies that do have some employees checking the sites rely in large part on their computer programs to filter and block the sites. All are doing their best to meet their customers' needs, blocking and unblocking sites as they find out about them.

Perhaps the best way to demonstrate the strengths and weaknesses of these programs is to show you some pages from our test of X-Stop's "librarian shadow." X-Stop says it has figured out how to block "illegal pornography" on the Web using the Miller test and an intelligent agent.

Consider the six tests which follow. Test 1 tried to access typical sexually-explicit sites. All sites were blocked. Test 2 attempted to access sites dealing with sex education. All sites were able to be accessed. Test 3 sought to access sites dealing with safe sex information. Here there was a mixed response. Notice, for instance, Test 4. Here we were able to access the site for "I might be a lesbian, what do I do?" and got the information on the following pages. In Test 5, we clicked on "I might be gay, what do I do?" and were blocked. The next page is the message sent when a site is blocked. The pages that follow Test 5 were accessed from a different browser and show the information and resource list which was blocked. Test 6 shows a safe sex site that was not blocked, but which contained an animation of a nude man putting a condom on his erect penis. WARNING: this page might be offensive to some.

Certainly, one should not make much of this test. We include it only for instructive purposes for those who have not tested filtering software themselves. As Robert Harrington puts it: "Putting aside the polemics and lawyer-talk, this debate comes down to these issues: those who favor restrictions face an increased danger of having their children exposed to pornography; opponents of restrictions face loss of some non-pornographic material blocked by mistake. Thus we have protection from pornography for minors on the one hand versus loss of some marginal information on the other." The readers of this report will need to judge whether this is an example of marginal information or not.

Back to the Table of Contents or voice your opinion.