|1.0 - 1.14||Paul Resnick,
University of Michigan
|last revised 04-Aug-99|
Status of This Document
This Frequently Asked Question (FAQ) white paper answers common questions about PICS in the context of intellectual freedom and censorship. It has no techincal standing what-so-ever, such information is found in the specifications. This document has no official W3C policy standing beyond describing the position the W3C has taken to date.
Please send questions to firstname.lastname@example.org .
The published articles describing PICS (Communications of the ACM, Scientific American) have focused on individual controls over the materials that are received on a computer. While those articles also mention the possibility of more centralized controls (e.g., by employers or governments), they describe only briefly the technical details and the intellectual freedom implications of such centralized controls. The civil liberties community has raised some alarms about those intellectual freedom implications. The goals for this Frequently Asked Questions (FAQ) document are to:
In 1995, policies were proposed in several countries, including the USA, to restrict the distribution of certain kinds of material over the Internet. In many but not all cases, protection of children was the stated goal for such policies (see, for example, CIEC: Citizens Internet Empowerment Coalition).
The focus on restricting inappropriate materials at their source is not well suited to the international nature of the Internet, where an information source may be in a different legal jurisdiction than the recipient. Moreover, materials may be legal and appropriate for some recipients but not others, so that any decision about whether to block at the source will be incorrect for some audiences.
PICS, the Platform for Internet Content Selection, is a set of technical specifications that facilitate recipient-centered controls on Internet content, rather than sender-centered controls. The following diagram illustrates recipient-centered controls:
Filtering software sits between a child (or any Internet user) and the available content. It allows access to some materials, and blocks access to other materials. Some filtering software directly analyzes content, typically looking for particular keywords. This FAQ, however, does not deal with that kind of software; it deals, instead, with filtering software that decides what to allow and what to block based on two information sources.
PICS was not the first technology based on the idea of recipient-centered controls. For example, SurfWatch was already on the market in the summer of 1995 when PICS development began. It is based on a particularly simple set of labels: a list of URLs to avoid. As another example, some firewalls that corporations had introduced for security purposes blocked access to certain IP addresses. PICS provides a set of technical specifications so that pieces of the picture could be provided by different entities, yet still work together.
The first and most important distinction that PICS introduced is a separation between labeling and filtering. A label describes the content of something. A filter makes the content inaccessible to some audience. While both labeling and filtering may introduce social concerns, the concerns are somewhat different. More generally, there are six roles that could all be filled by different entities:
PICS itself actually fills none of the six roles listed above! PICS is a set of technical specifications that makes it possible for these roles to be played by independent entities.
For example, RSACi and SafeSurf have each defined labeling vocabulary and criteria for rating. They each wrote down a vocabulary in a machine-readable format that PICS specifies. RSACi has four categories in its vocabulary, language, nudity, sex, and violence; SafeSurf has more categories. Because they write down their vocabularies in the PICS format, label distribution software (e.g., from IBM and Net Shepherd) and filtering software (e.g., from Microsoft, IBM, and others) can process labels based on those vocabularies. Even though RSACi and SafeSurf have each specified a labeling vocabulary and criteria for assigning labels, neither of them actually assigns labels: they leave it up to the authors of documents to apply to criteria to their own documents, or self-label as PICS documents call it. Other services, such as CyberPatrol and Net Shepherd, take on both of the first two roles, choosing the labeling vocabulary and employing people to actually assign labels. The PICSRules specification provides a common format for expressing filtering criteria, which makes it easy for one entity to set filtering criteria which are then installed and run by someone else. For example, a parent might, with one mouse click install filtering criteria suggested by some trusted organization, say a local church group, even though that organization provides neither rating labels nor the filtering software.
Yes. While the motivation for PICS was concern over children accessing inappropriate materials, it is a general "meta-data" system, meaning that labels can provide any kind of descriptive information about Internet materials. For example, a labeling vocabulary could indicate the literary quality of an item rather than its appropriateness for children. Most immediately, PICS labels could help in finding particularly desirable materials (see, for example, NetShepherd's label-informed Alta Vista search), and this is the main motivation for the ongoing work on a next generation label format that can include arbitrary text strings. More generally, the W3C is working to extend Web meta-data capabilities generally and is applying them specifically in the following projects:
Regardless of content control, meta-data systems such as PICS are going to be an important part of the Web, because they enable more sophisticated commerce (build and manage trust relationships), communication, indexing, and searching services.
"The promise of digital commerce is that it will allow you to use the Internet to purchase the services of the best organic gardening advisors or mad cow disease specialists, whether they live in Santa Clara or Timbuktu. To do this, you need to do more than verify that the person at the other end of the wire is who he says he is. You need to assess competence, reliability, judgment. In other words, you need a system of branding, but applied much more widely for highly specialized and hard-to-evaluate services and products. You need value-added services that will not only lead you to the right product or service but also rate its quality or otherwise vouch for it."
(Forbes ASAP 12/96 p 69)
This seemingly straightforward question, upon closer inspection, turns out to be many different questions when asked by different people. Many people are concerned about governments assuming one or more of the roles described in the answer to the previous question. Others are concerned about employers setting filtering rules, abuse of power by independent labelers, or a chilling effect on speech even if speech is not banned outright. People also employ different definitions of censorship. The most expansive definition is, "any action by one person that makes otherwise available information unavailable to another person." Under this expansive definition, even a parent setting filtering rules for a child would count as censorship. PICS documents have adopted the more restrictive definition of censorship as actions that limit what an individual can distribute, and use the term "access controls" for restrictions on what individuals can receive. But the distinction blurs if a central authority restricts access for a set of people. Finally, people have different definitions of "enable." Some would say that PICS enables any application that uses PICS-compatible components, while we reserve the term "enables" for applications that can easily be implemented with PICS-compatible components but could not be easily implemented otherwise.
Given the variety of implicit questions, it doesn't make sense to provide a blanket answer to the question of whether PICS enables censorship. This FAQ answers many of the specific questions that people often mean when they ask the more general question. For example, we ask questions about whether PICS makes it easier or harder for governments to impose labeling and filtering requirements. If you believe there's another specific question that should be addressed, please send it to email@example.com, for possible inclusion in a later version.
Yes. A government could try to assume any or all of the six roles described above, although some controls might be harder than others to enforce. As described below, governments could assume some of these roles even without PICS, while other roles would be harder to assume if PICS had not been introduced. It's important to note that W3C does not endorse any particular government policy. The purpose of this FAQ is to explain the range of potential policies and to explore some of the impacts of those policies on both the climate of intellectual freedom and the technical infrastructure of the World Wide Web. Potential government policies:
Yes; all other things being equal, when the price of something drops, more of it will be consumed.
Yes; for example, a national proxy-server/firewall combination that blocks access to a government-provided list of prohibited sites does not depend on interoperability of labels and filters provided by different organizations. While such a setup could use PICS-compatible technology, a proprietary technology provided by a single vendor would be just as effective. Other controls, based on individual or local choices, benefit more from mixing and matching filtering software and labels that come from different sources, which PICS enables. Thus, there should be some substitution of individual or local controls for centralized controls, although it is not obvious how strong this substitution effect will be. In Europe initial calls for centralized controls gave way to government reports calling for greater reliance on individual recipient controls; the end results of these political processes, however, are yet to be determined.
An IP address identifies the location of a computer on the Internet. A URL identifies the location of a document. To simplify a little, a URL has the form http://<domain-name>/<filename>. A web browser first resolves (translates) the domain-name into an IP address. It then contacts the computer at that address and asks it to send the particular filename. Thus, a label that applies to an IP address is a very broad label: it applies to every document that can be retrieved from that machine. Labeling of URLs permits more flexibility: different documents or directories of documents can be given different labels.
This difference of granularity will, naturally, have an impact on filtering. Filters based on IP addresses will be cruder: if some but not all of the documents available at a particular IP address are undesirable, the filter will have to either block all or none of those documents. PICS, by contrast, permits labeling of individual URLs, and hence permits finer grain filters as well.
Yes. Without a common format for labels, authors could not label themselves in a way that filtering programs could make use of. PICS provides that format.
It enables such a requirement to have more impact. A government requirement of self-labeling would have little impact if the labels were not usable by filtering programs. PICS provides the common format so that filtering software from one source can use labels provided by other sources (authors in this case).
Although universal agreement is not necessary, there does need to be some harmonization of vocabulary and labeling criteria, so that labels provided by different authors can be meaningfully compared.
Yes. PICS provides a language-independent format for expressing labels. If governments agreed on a common set of criteria for assigning labels, the criteria could be expressed in multiple languages, yet still be used to generate labels that can be compared to each other.
Both kinds of labeling could be effective, but only if a high percentage of the materials of a particular type are labeled. If the inappropriate materials are labeled, then a filter can block access to the labeled items. If the appropriate materials are labeled, then a filter can block access to all the unlabeled items.
Yes. Anyone can create a PICS label that describes any URL, and then distribute that label to anyone who wants to use that label. This is analogous to someone publishing a review of your web site in a newspaper or magazine.
If a lot of people use a particular organization's labels for filtering, that organization will indeed wield a lot of power. Such an organization could, for example, arbitrarily assign negative labels to materials from its commercial or political competitors. The most effective way to combat this danger is to carefully monitor the practices of labeling services, and to ensure diversity in the marketplace for such services, so that consumers can stop using services that abuse their power.
PICS documents use the term "label" broadly to refer to any machine-readable information that describes other information. Even information that merely classifies materials by topic or author (traditional card catalog information) would qualify as labels if expressed in a machine-readable format. The PICS developers recognized that the term "label" has a narrower meaning, with negative connotations, for librarians and some other audiences, but it was the most generic term the PICS creators could find without reverting to technical jargon like "metadata."
In media with centralized distribution channels, such as movies, labeling and filtering are not easily separated. For example, unrated movies are simply not shown in many theaters in the USA. In addition to its technical contribution, PICS makes an intellectual contribution by more clearly separating the ideas of labeling and filtering. Many of the negative connotations associated with "labeling" really should be associated with centralized filtering instead. There are, however, some subtle questions about the impact of labeling itself, as articulated in the next two questions.
Matt Blaze (personal communication) describes this concern with an analogy to discussions at local school board meeting about books to be read in a high school English class. Ideally, the discussion about a particular book should focus on the contents of the book, and not on the contents of a review of the book, or, worse yet, a label that says the book contains undesirable words.
There will always be a tradeoff, however, between speed of decision-making and the ability to take into account subtleties and context. When a large number of decisions need to be made in a short time, some will have to be made based on less than full information. The challenge for society, then, will be to choose carefully which decisions merit full discussion, in which case labels should be irrelevant, and which decisions can be left to the imperfect summary information that a label can provide. The following excerpt from Filtering the Internet summarizes this concern and the need for eternal vigilance:
"Another concern is that even without central censorship, any widely adopted vocabulary will encourage people to make lazy decisions that do not reflect their values. Today many parents who may not agree with the criteria used to assign movie ratings still forbid their children to see movies rated PG-13 or R; it is too hard for them to weigh the merits of each movie by themselves.
Labeling organizations must choose vocabularies carefully to match the criteria that most people care about, but even so, no single vocabulary can serve everyone's needs. Labels concerned only with rating the level of sexual content at a site will be of no use to someone concerned about hate speech. And no labeling system is a full substitute for a thorough and thoughtful evaluation: movie reviews in a newspaper can be far more enlightening than any set of predefined codes."
This is indeed a serious concern, explored in detail by Jonathan Weinberg in his law review article, Rating the Net. The following excerpt from Filtering the Internet acknowledges that materials of limited appeal may not reach even the audiences they would appeal to, but argues that labeling is merely a symptom rather than a cause of this underlying problem:
"Perhaps most troubling is the suggestion that any labeling system, no matter how well conceived and executed, will tend to stifle noncommercial communication. Labeling requires human time and energy; many sites of limited interest will probably go unlabeled. Because of safety concerns, some people will block access to materials that are unlabeled or whose labels are untrusted. For such people, the Internet will function more like broadcasting, providing access only to sites with sufficient mass-market appeal to merit the cost of labeling.
While lamentable, this problem is an inherent one that is not caused by labeling. In any medium, people tend to avoid the unknown when there are risks involved, and it is far easier to get information about material that is of wide interest than about items that appeal to a small audience."
PICSRules is a language for expressing filtering rules (profiles) that allow or block access to URLs based on PICS labels that describe those URLs. The purposes for a common profile-specification language are:
Sharing and installation of profiles. Sophisticated profiles
may be difficult for end-users to specify, even through
well-crafted user interfaces. An organization can create a recommended profile for children of a certain age. Users who
trust that organization can install the profile rather than specifying one from scratch. It will be easy to change the active
profile on a single computer, or to carry a profile to a new computer.
Communication to agents, search engines, proxies, or other servers. Servers of various kinds may wish to tailor
their output to better meet users' preferences, as expressed in a profile. For example, a search service can return only
links that match a user's profile, which may specify criteria based on quality, privacy, age suitability, or the safety of
Portability betwen filtering products. The same profile will work with any PICSRules-compatible product.
No, but an effective national firewall would make it possible for a government to impose PICS-based filtering rules (or non PICS-based filtering rules) on its citizens. A firewall partitions a network into two components and imposes rules about what information flow between the two components. The goal of a national firewall is to put all the computers in the country into one component, and all computers outside the country into the other component. This is difficult to do, especially if people deliberately try to find out connections (e.g., telephone lines) between computers inside the country and those outside the country. Given a successful partition, however, PICS could be used to implement the filtering rules for a firewall. In particular, the government could identify prohibited sites outside the country that people inside the country could not access; such a filtering could be implemented based on PICS-formatted labels or, without relying on PICS-compatible technology, with a simple list of prohibited URLs.
No. PICSRules can provide a way to express filtering preferences, but has no impact on the ability of a government to partition the computers inside a country from those outside the country.
ISP compliance with government prohibition lists is already practical, even without
PICS. It would also be possible to comply using PICS-based technologies. PICS does make it
easier for ISPs to comply with a government requirement to block access to sites labeled
by non-governmental entities (including those that are self-labeled by the authors of the
Governments can make such requirements with or without PICSRules. PICSRules does make it possible for governments to precisely state filtering requirements, and perhaps simplify ISP compliance with changing government requirements, if the ISP implements a general interpreter for the PICSRules language.
Yes. PICS enables mixing and matching of the five roles. In particular, a service provider could install and run filtering software on a proxy server, but allow individuals to choose what filtering rules will be executed for each account. AOL already offers a primitive version of this idea, not based on PICS; parents can turn the preset filtering rules on or off for each member of the family.
No. Governments could require the use of filters on clients. The city of Boston, for example, requires public schools to install a client-based filtering product on all computers with Internet access, and requires public libraries to install a client-based filtering product on all computers designated for children.
W3C leaves this question to the political and legal processes of each country. Some people argue that unrestricted access to information is a fundamental human rights question that transcends national sovereignty. W3C has not adopted that position.
W3C leaves this question to the political and legal processes of each country.
W3C's mission is to "realize the full potential of the Web." The following two points are taken from a talk by Jim Miller at the WWW6 conference:
Thus, for example, when discussing the CDA-type legislation with government officials in the U.S. or abroad, it is appropriate for W3C to point out that sender-based restrictions are not likely to be effective at keeping all materials of a particular kind away from children, and that there could be unintended consequences in terms of chilling free speech or keeping the Web from reaching its full potential as a medium for communication and cultural exchange. W3C does not, however, debate with government officials about their perceived policy requirements. For example, Germany has a policy requirement of restricting access to hate speech while the U.S. does not: W3C does not try to convince either country that the other country's choice of policy requirements is better.
Some people(see The Net Labeling Delusion) have criticized the following paragraph from the CACM article on PICS:
Not everyone needs to block reception of the same materials. Parents may not wish to expose their children to sexual or violentimages. Businesses may want to prevent their employees from visiting recreational sites during hours of peak network usage. Governments may want to restrict reception of materials that are legal in other countries but not in their own. The off button (or disconnecting from the entire Net) is too crude: there should be some way to block only the inappropriate material. Appropriateness, however, is neither an objective nor a universal measure. It depends on at least three factors:
1. The supervisor: parenting styles differ, as do philosophies of management and
2. The recipient: whats appropriate for one fifteen year old may not be for an eight-year-old, or even all fifteen-year-olds.
3. The context: a game or chat room that is appropriate to access at home may be inappropriate at work or school.
The main point of this section is to underscore the fact that people disagree about what materials are appropriate in what contexts. This point is illustrated at several levels of granularity: invidual children, organizations, and governments. The critcism focuses on the mention of possible government blocking, which did not appear in an earlier draft of the paper. We believe the example about differences in laws between countries is useful in explaining why there is a need for flexible, receiver-based controls rather than the kind of sender-based controls (e.g., the CDA) that most policy discussions were focusing on at the time.
The objection to the use of this example rests on an argument that governments should never designate any content as illegal. That argument is not widely accepted (in the U.S., for example, "obscenity" laws have been deemed constitutional, even though the CDA's "indecency" provisions were not). A more widely held position is that governments should not restrict political materials as a means of controlling their citizens. W3C leaves discussions about which materials should be illegal in a particular country to the political realm rather than the technological realm. W3C does, however, point out to policy makers, however, that it's not necessary to make materials illegal if they are offensive to some people but not others: end-user controls are a more flexible method of handling such materials.
Licensing such a technology was not considered to be a feasible option during the time of the CDA. Not only would it have undercut the "neutrality" and appeal of the technology, the W3C then would have had to be in the position of determining who should and should not use it; this is not a role the W3C is competent to play.
Yes. W3C is pleased that IBM has introduced a proxy server that can filter based on PICS labels, and encourages the development of other PICS-compatible servers. As discussed above, filter processing can be centralized at a proxy server while still permitting individuals to choose the filtering rules.
In addition to acting in the political arena, it would probably be helpful to implement positive uses of labels, such as searching applications. It is surpassingly difficult for people unfamiliar with computers to imagine new applications. By building prototypes and demonstrating them, it may be possible to focus policy-makers' energies on those uses of technology that accord with your political values.