fight it through legal means. All that free speech activists could then do is write powerful, but largely invisible, articles like the ACLU’s famous plea.
It has taken key civil rights organizations too long to recognize this private threat to free-speech values. The tradition of civil rights is focused directly on government action alone. I would be the last to say that there’s not great danger from government misbehavior. But there is also danger to free speech from private misbehavior. An obsessive refusal to even consider the one threat against the other does not serve the values promoted by the First Amendment.
But then what about public filtering technologies, like PICS? Wouldn’t PICS be a solution that avoided the “secret list problem” you identified?
PICS is an acronym for the World Wide Web Consortium’s Platform for Internet Content Selection. We have already seen a relative (actually, a child) of PICS in the chapter about privacy: P3P. Like PICS, is a protocol for rating and filtering content on the Net. In the context of privacy, the content was made up of assertions about privacy practices, and the regime was designed to help individuals negotiate those practices.
With online speech the idea is much the same. PICS divides the problem of filtering into two parts — labeling (rating content) and then filtering (blocking content on the basis of the rating). The idea was that software authors would compete to write software that could filter according to the ratings; content providers and rating organizations would compete to rate content. Users would then pick their filtering software and rating system. If you wanted the ratings of the Christian Right, for example, you could select its rating system; if I wanted the ratings of the Atheist Left, I could select that. By picking our raters, we would pick the content we wanted the software to filter.
This regime requires a few assumptions. First, software manufacturers would have to write the code necessary to filter the material. (This has already been done in some major browsers). Second, rating organizations would actively have to rate the Net. This, of course, would be no simple task; organizations have not risen to the challenge of billions of web pages. Third, organizations that rated the Net in a way that allowed for a simple translation from one rating system to another would have a competitive advantage over other raters. They could, for example, sell a rating system to the government of Taiwan and then easily develop a slightly different rating system for the “government” of IBM.
If all three assumptions held true, any number of ratings could be applied to the Net. As envisioned by its authors, PICS would be neutral among ratings and neutral among filters; the system would simply provide a language with which content on the Net could be rated, and with which decisions about how to use that rated material could be made from machine to machine[48].
Neutrality sounds like a good thing. It sounds like an idea that policymakers should embrace. Your speech is not my speech; we are both free to speak and listen as we want. We should establish regimes that protect that freedom, and PICS seems to be just such a regime.
But PICS contains more “neutrality” than we might like. PICS is not just horizontally neutral — allowing individuals to choose from a range of rating systems the one he or she wants; PICS is also vertically neutral — allowing the filter to be imposed at any level in the distributional chain. Most people who first endorsed the system imagined the PICS filter sitting on a user’s computer, filtering according to the desires of that individual. But nothing in the design of PICS prevents organizations that provide access to the Net from filtering content as well. Filtering can occur at any level in the distributional chain — the user, the company through which the user gains access, the ISP, or even the jurisdiction within which the user lives. Nothing in the design of PICS, that is, requires that such filters announce themselves. Filtering in an architecture like PICS can be invisible. Indeed, in some of its implementations invisibility is part of its design[49].
This should set off alarms for those keen to protect First Amendment values — even though the protocol is totally private. As a (perhaps) unintended consequence, the PICS regime not only enables nontransparent filtering but, by producing a market in filtering technology, engenders filters for much more than Ginsberg speech. That, of course, was the ACLU’s legitimate complaint against the original CDA. But here the market, whose tastes are the tastes of the community, facilitates the filtering. Built into the filter are the norms of a community, which are broader than the narrow filter of Ginsberg. The filtering system can expand as broadly as the users want, or as far upstream as sources want.
The H2M+KMB solution alternative is much narrower. It enables a kind of private zoning of speech. But there would be no incentive for speakers to block out listeners; the incentive of a speaker is to have more, not fewer, listeners. The only requirements to filter out listeners would be those that may constitutionally be imposed — Ginsberg speech requirements. Since they would be imposed by the state, these requirements could be tested against the Constitution, and if the state were found to have reached too far, it could be checked.
The difference between these two solutions, then, is in the generalizability of the regimes. The filtering regime would establish an architecture that could be used to filter any kind of speech, and the desires for filtering then could be expected to reach beyond a constitutional minimum; the zoning regime would establish an architecture for blocking that would not have this more general purpose.
Which regime should we prefer?
Notice the values implicit in each regime. Both are general solutions to particular problems. The filtering regime does not limit itself to Ginsberg speech; it can be used to rate, and filter, any Internet content. And the zoning regime, in principle, is not limited to zoning only for Ginsberg speech. The <H2M> kids-ID zoning solution could be used to advance other child protective schemes. Thus, both have applications far beyond the specifics of porn on the Net.
At least in principle. We should be asking, however, what incentives are there to extend the solution beyond the problem. And what resistance is there to such extensions?
Here we begin to see the important difference between the two regimes. When your access is blocked because of a certificate you are holding, you want to know why. When you are told you cannot enter a certain site, the claim to exclude is checked at least by the person being excluded. Sometimes the exclusion is justified, but when it is not, it can be challenged. Zoning, then, builds into itself a system for its own limitation. A site cannot block someone from the site without that individual knowing it[50] .
Filtering is different. If you cannot see the content, you cannot know what is being blocked. Content could be filtered by a PICS filter somewhere upstream and you would not necessarily know this was happening. Nothing in the PICS design requires truth in blocking in the way that the zoning solution does. Thus, upstream filtering becomes easier, less transparent, and less costly with PICS.
This effect is even clearer if we take apart the components of the filtering process. Recall the two elements of filtering solutions — labeling content, and then blocking based on that labeling. We might well argue that the labeling is the more dangerous of the two elements. If content is labeled, then it is possible to monitor who gets what without even blocking access. That might well raise greater concerns than blocking, since blocking at least puts the user on notice.
