Congress' Censorware Boondoggle

Congress, as part of the budget bill which passed at years' end, has saddled schools and libraries with the Children's Internet Protection Act (CIPA). This law applies to any school or library accepting E-rate funding for Internet connectivity. The E-rate funds are derived from a tax on your telephone calls.

CIPA is a classic legislative boondoggle, as it prescribes a solution which does not remotely address the problem it identifies. The purpose of CIPA is to protect children against Internet material harmful to minors, and (despite its name) to protect adults against obscene material. It does so by mandating that schools and libraries accepting e-rate funding install a "technology protection measure"--blocking software or "censorware".

In Sex, Laws and Cyberspace, a book I co-authored with Mark Mangan (Henry Holt 1996), we naively endorsed censorware as a "less restrictive alternative" to government censorship such as the Communications Decency Act. Within six months after the book first came out in 1996, I discovered that < the web site we had put up describing it was blocked by several censorware products--undoubtedly because of the word "sex" in the title. The November 1995 issue of my monthly newsletter, The Ethical Spectacle, containing essays about the views of Catherine Mackinnon, Wendy McElroy and Nadine Strossen, was blocked because of the word "Pornography" in the title.

In 1996, the massive over-blocking of all censorware products then on the market began to come to public attention. For example, the X-Stop product, which claimed to block obscene material only, blacklisted a Quaker web site, the American Association of University Women, and the Aids Quilt. Soon after, a group of concerned people formed the Censorware Project, with the express mission of fighting the use of censorware in public libraries, and to spotlight the amazing overbreadth of these products' blacklists. The group analyzed and released reports on products such as Bess, Websense, and Cyberpatrol. In a Censorware Project report entitled Protecting Judges Against Liza Minelli, we noted that Websense, which had been installed on the federal courts' computer network, blocked a Liza Minelli fan page. The Cyberpatrol product repeatedly blacklisted a site called Maplesoccer.org, which consisted of the schedule for a local youth soccer organization.

All censorware products share inevitable and very serious flaws based on the difficulty of identifying material "worthy" of being blocked. While any individual parent could certainly make up a blacklist based on his or her personal values, censorware companies take on the task of making universal "one size fits all" blacklists for children of every age and cultural and political background. While this raises serious practical issues about the efficacy of these products for individual families who may not share the views of the company compiling the blacklist, the problem becomes much more severe when these products are tapped-- as CIPA has done--for the role of policing only "illegal" material. (I'll return to this issue in a moment.)

Secondly, censorware companies are completely inadequate to the task they undertake to perform, of policing the entire Web. Though no-one knows exactly how large the World Wide Web is today, a February 1999 article in Nature Magazine concluded the Web already contained 800 million pages of information. You don't have to be a math wizard to calculate the amount of time it would take a censorware company staff of even one hundred reviewers to get through all those pages. All censorware companies are extraordinarily secretive about the number of people they use for Web review, so the teams are possibly far smaller than 100 people. Censorware companies are relatively small businesses; N2H2, Inc. of Seattle, which distributes the Bess product, reported that it had $2.5 million in revenue in first quarter of fiscal 2001, giving it a run rate of only ten million dollars for the year. N2H2 is one of only two censorware companies which are publicly traded; others are not required to report their results, but are unlikely to be larger than the censorware companies which have already achieved a public offering. A company with only ten million dollars in annual revenue would certainly not be able to field a very substantial staff of reviewers.

Because of the impossibility of keeping up with the ever-changing web with a small group of human reviewers, there is evidence that most censorware companies also use technological means to identify pages for blocking--so-called "spider" programs which crawl the Web looking for particular keywords. Though most censorware companies have made the claim that no page is ever added to the blacklist without human review, it is improbable that anyone looked at innocuous sites like Maplesoccer.org, the Quaker or Liza Minelli pages, before blacklisting them. The censorware companies' poor record of human review, coupled with the use of these stupid and literal-minded spiders, results in an astonishingly high rate of error.

While the censorware companies respond to every new revelation by claiming they are cleaning up their act (or simply deny the blocking had ever occurred) the Censorware Project found across three years of evaluations that things never got any better. The companies removed from the blacklist the sites we called to their attention, and went blithely on to blacklist new innocuous sites--because, with their flawed process, they simply could not do any better. In some cases, we found that the same sites they unblocked were repeatedly added back to the blacklist, like maplesoccer.org, which was unblocked and re-added at least three times. When we would go back a year after last examining a product and look at it again, we always discovered that the error rate remained as high as before.

Although the Censorware Project's main issue was over-blocking and the resulting censorship of innocuous and socially useful sites, when we looked for it we naturally found substantial underblocking as well. Every product we examined let pass substantial numbers of hard core sex sites which the human reviewers, or the spider software, had not yet found. Again, the problem of reviewing 800 million plus web pages is insurmountable by any means within the budget of a relatively small for profit Internet company.

In 1998, a group named Mainstream Loudoun brought suit against the library system of Loudoun County, Virginia, which had just voted to install the X-Stop censorware product. (I became a plaintiff in that case as part of a group of Web site operators, represented by the ACLU, whose sites had been blocked by X-Stop.) Federal Judge Leonie Brinkema, a former librarian, issued an injunction banning X-Stop's use in the library precisely because of the over-breadth, citing many of the sites we had identified in the X-Stop Files article. (The court's two rulings can be found at http://www.pfaw.org/courts/loudoun_opinion.shtml and http://www.pfaw.org/courts/loudoun_ruling.shtml.) Judge Brinkema held that the Supreme Court's ruling in Board of Education v. Pico, 457 U.S. 853 (1982) governed. In Pico, the Supreme Court had held that a local school board could not descend upon a high school library and remove works by authors such as John Updike which offended its conservative sensibilities.

Censorware has been the darling of the fundamentalist right since the defeat of the Communications Decency Act; groups like the American Family Association and the National Law Center for Children and Families have been championing government-mandated censorware for years. This coalition, which includes people who have campaigned in the past for the banning of Huckleberry Finn from libraries, has strenuously advanced the argument that censorware does not remove anything from the library, as the school board did in Pico; it simply helps the librarian choose not to include material, as a librarian would be entitled to choose not to purchase Madonna's Sex (a favorite example). Judge Brinkema held to the contrary. Though no library is constitutionally required to offer Internet access, once it decides to do so, Brinkema analogized the Internet to a huge encyclopedia purchased by the library. Then, along come the X-stop censors with their scissors, she said, offering to "redact" the encyclopedia.

The local library board chose not to appeal the decision, so there was no opportunity to obtain affirmation from the Supreme Court. The decision, still the only one on the question of whether libraries can adopt censorware, has been disregarded by censorware proponents and by Congress, who hope that federal courts elsewhere, and the Supreme Court when a case finally gets up to them, would rule differently.

A member of the American Family Association might, of course, feel warmly towards a censorware product which he believed embodied his values (such as Cybersitter, which blocked the National Organization of Women site as a "radical lesbian" page). The problem of tailoring censorware products to meet defined legal standards (rather than the putative AFA member's subjective values) is much more difficult.

CIPA mandates that schools and libraries accepting e-rate funding use censorware to protect adults from obscene material. The Supreme Court in Miller v. California defined obscenity as "prurient, patently offensive" material lacking SLAP (scientific, literary, artistic or political) value. The problem, of course, is that censorware reviewers, acting under secret criteria, are in the position of deciding that material is legally obscene which has not been ruled so by a court. Further complicating the problem is the fact that the Miller case adopted local community standards of "prurience" and "patent offensiveness", meaning that speech which is acceptable in California may be obscene in Tennessee (exactly the result in the Amateur Action bulletin board case we wrote about in the first chapter of Sex, Laws and Cyberspace). No censorware product has ever yet attempted to create fifty different state versions (let alone mirror the differing community standards which may occur within a state.) Thus, the reviewing staff of censorware companies, which tends to consist of poorly-paid students and homemakers looking for a little extra income, is now in the position of making determinations best left to judges and First Amendment lawyers. All censorware companies keep secret their criteria for choosing sites for the blacklist, so it is impossible to tell if, or how, they are applying Miller in making their decisions. No censorware company has ever even claimed to have a setting which blocks only obscenity, since X-Stop met embarrassment in 1997. It was the "Obscenity Load" version of X-Stop which blocked the Quaker pages, the American Association of University Women,and the AIDS Quilt.

The use of spiders to identify sites for blocking exacerbates the problem. Deciding which material is obscene or is harmful to minors is an extremely subjective human endeavor. No "artificial intelligence" program has ever been created which can do this; the job is difficult enough for a human jury to perform. In fact, though the censorware companies are also highly secretive about their technology, every indication is that the spiders are not sophisticated attempts at rule-based decision engines, but simple search programs which prowl the net looking for the occurrence of keywords like "sex" or "pornography".

Also consider that most censorware products are advertised as created specifically for the protection of children, yet CIPA mandates that these same products be used to block sites from adult view. There is a long line of Supreme Court cases holding that adults cannot be reduced to reading or viewing only that material which is fit for children. In her decision, judge Brinkema held that the library's use of X-Stop was "overinclusive because...it limits the access of all patrons, adult and juvenile, to material deemed fit for juveniles."

A closely related problem is that government is not permitted to delegate censorship decisions to private groups. A long line of cases in various states have unanimously held that state governments cannot make criminal laws adopting the MPAA movie ratings (NC17, R, PG). The MPAA's private efforts in categorizing films are simply too imprecise and arbitrary to be utilized by government. (For a discussion of these and other relevant cases, see my brief Purchase of Blocking Software by Public Libraries is Unconstitutional. The same is abundantly true of censorware companies, which refuse even to open their blacklists to the libraries or schools who will be using them. Rude as it is to type in capitals in an Internet communication: CIPA FORCES LIBRARIES TO ADOPT BLACKLISTS OF UNKNOWN CONTENT, COMPILED BY UNSKILLED CENSORS, USING AN UNREVEALED METHODOLOGY.

Congress has foisted a technology which does not work upon schools and libraries. Lie detector tests have existed for many decades, but their use as evidence in criminal trials has never been permitted because they are just not scientifically reliable enough. You wouldn't want Congress to mandate the use of lie detector tests in every criminal trial. And you certainly would expect Congress, as a group of responsible legislators, to take a close look at the efficacy of any technology before requiring its use.

Congress didn't do its homework here. The truly remarkable aspect of CIPA is the extent to which Congress relied on the shaky, easily disproved claims of the censorware companies, without doing any due diligence. Any librarian who has dealt with censorware knows the truth. A technical coordinator at one school using the Bess censorware product wrote that teachers in his building were becoming so frustrated with Bess that "they decided it wasn't worth the hassle to use the Web"--a common response from teachers when they first discover that students are unable to view sites they have been assigned for homework. Because Bess, a product of the publicly traded N2H2 Inc., is typically installed on a server serving the schools of an entire district, the censorware had to be turned off for the entire district every time the students in one classroom wanted to access an erroneously blocked site they needed for a class assignment.

Did Congress really believe that censorware works? CIPA remarkably contains a provision that requires a study, within eighteen months after its passage, to determine whether censorware is effective for the job that CIPA has tapped it to do. Congress said, let's mandate the use of the technology now, and determine whether it all works later.

The fundamentalists are smiling, because they have achieved their real agenda. It won't stand up in court, but what they did was clever. They wanted much broader censorship; a law blocking only legally obscene material wouldn't serve their purposes, because they believe much First Amendment-protected material is destructive of children (like Huckleberry Finn). So they wrote a law requiring the blocking only of obscene material, but requiring the use of technology incapable of doing only that. Which will inevitably blacklist much more than that.

CIPA is a boondoggle, a huge waste of public money and time on technology that will not serve the stated purpose, but will inflict a lot of harm along the way.