Content Management Tools: Filters, Labels ...

Blocking access to online content - whether you're a corporation, the Taliban or a concerned parent - has become big business, with over 100 vendors in the US alone. This page explores filters and other content management technologies.

     principles

Despite hype from promoters, artificial intelligence is insufficiently advanced to determine 'on the fly' whether a graphic is offensive and thus prevent its display online. Most censorship technologies instead attempt to block access to specific sites, generally using a labelling system.

Sites are examined, manually or automatically, to determine whether they should be labelled. 

That examination encompasses whether the domain name or metadata includes terms deemed offensive, a process that's problematical since many sites do not have extensive metadata. It also includes scrutiny of whether text, graphics, audio or video within the site meet the examiner's criteria. 

'Objectionable' sites/pages are then labelled, either on the site by its owner or in an independent list. Browsers can be modified to recognise those labels and thus restrict access to specific sites.

Practice is, alas, more problematical. There is no global requirement that sites be labelled by their owners or agreement about labelling criteria. Operators of sites deemed offensive by a particular market or government generally respond by moving to a domain in a less restrictive jurisdiction or ignoring a rating.  

Understandably, as the number of pages on the web increases, manual rating is failing to keep pace with growth (eg probably covers less than 1% of domains). There's disagreement about whether past ratings have been kept up to date. 

Critics argue that while rating per se is not contentious, the way in which it's been administered by particular companies and associations suggests considerable scope for abuse. 

In Australia, for example, the sites of various parliaments, government agencies and advocacy groups have been blocked (the federal parliament site contains words such as 'whip'). 

Similar problems are evident in consumer 'seals' programs, discussed in our Consumers guide, and in past initiatives such as the US V-Chip plan for blocking offensive television broadcasts.

     PICS and ICRA

Many of the filters, blocking mechanisms and other content management regimes are based on the Platform for Internet Content Selection (PICS), a metadata-based standard for internet content.  

PICS was developed in association with the World Wide Web Consortium as part of that body's interest in the 'architecture' of the Internet. Despite W3C endorsement it has never really got off the ground. It provides for tagging of web pages, eg allows them to be labelled as containing violent or sexually-explicit material and thereby excludes access from particular browsers. It does not specify the nature of the labels or their derivation.

PICS is a building block for the Recreational Software Advisory Council (RSAC) rating scheme administered by the Internet Content Rating Association (ICRA), an industry body concerned with the invidious task of developing a viable content 'advisory' scheme, alerting surfers that there may be something unpleasant in the waters ahead.

ICRA's received some degree of endorsement from the EU, along with the inevitable denunciations from zealots who regard any content identification tool as tantamount to book burning. The 2000 report of the ICRA Advisory Board, drawing on the 'Best Practices' model (RTF) developed by the Information Society Project (ISP) at Yale's Law School, was construed by some as 'back to the drawing board'. In December 2000 ICRA released a more sophisticated rating framework with endorsement by the CDT, arguably a major step forward. In February 2001 that framework was extended to several languages other than English.

Lawrence Lessig's Tyranny In The Infrastructure article in WIRED is an assessment of PICS by one of the more hardheaded US legal theorists.

     technology studies

As noted earlier in this guide, content labelling is problematical: it's frequently hit and miss, it's consistently over-hyped, and endorsement by government in Australia and overseas is ill-founded.  

Filters lauded in Australia have excluded the most benign of sites while permitting access to those with 'explicit' content.  Studies of the "state-of-the-art" BAIR filter - endorsed alas by the Commonwealth government - in June 2000 demonstrated that while it excludes access to images of dogs, journalists, trees and vegetables it rates images of group and oral sex as acceptable fare for the kids.

The 1999 report Filters & Freedom by the Electronic Privacy Information Center (EPIC) is a useful starting point in understanding the technology. EPIC's site, like that of the Computer Professionals For Social Responsibility (CPSR) contains a range of information about 'censorware'. Another document of value is the COPA Commission's final report, which drew on the more detailed study on Risk & the Internet: Perception and Reality by Christopher Hunter & Eric Zimmer.

The latter reflected Hunter's earlier thesis - Filtering the Future?: Software Filters, Porn, PICS, and the Internet Content Conundrum - and paper on Internet Filter Effectiveness: Testing Over & Underinclusive Blocking Decisions of Four Popular Filters.

In June 2000 the EU released the final report of the Internet Content Rating For Europe (INCORE) project, exploring a pan-European content rating and filtering regime. The report was more cautious than material from the Australian government. 

The EU's Joint Research Centre is now engaged in a long-term project to benchmark filtering software and services, of particular significance given the dubious value of vendor statements and many of the government endorsements. 

Filtering the Internet: A Best Practices Model is a useful report from the Information Society Project at Yale Law School. There's a more comprehensive exploration in David Sobel's Filters & Freedom: Free Speech Perspectives on Internet Content Controls (Washington, Electronic Privacy Information Center 99).

Lawrence Lessig's paper What Things Regulate Speech: CDA 2.0 vs. Filtering and the 1998 paper by Lessig & Paul Resnick on Zoning Speech on the Internet: A Legal and Technical Model are both recommended. 

The Censorware Project, to the left of Lessig and bodies such as the CDT, has produced a number of reports on specific filters such as Bess, Cyberpatrol, X-Stop, NetNanny, CyberSitter, Smartfilter and Websense. 

Peacefire, a feisty libertarian group, has useful reports on technical aspects of filters - underwhelming and overhyped - and how they're implemented, eg noting that CyberSitter blocked the TIME magazine site after the publisher criticised its policies. 

In contrast Filtering Facts (FF) is a web site in support of filtering; arguably the emphasis is on filtering, less on the facts.

     in Australia

The ABA's requirements for restricted access systems are available on its site. That site also features the list of Australian government "approved filters".  Australian ISPs are required to offer their customers a blocking program from an approved list of 16 products as part of the Approved Code of Practice. 

Curiously, the ABA's criteria for choosing the filters do not include whether they work or not. An ABA spokesperson told us that the emphasis was on whether the software was easy to load rather than whether it performed as required. We consider that a more nuanced and informed policy - embracing user education, self-regulation and even common sense - would better achieve the government's objectives.

The filtering regime reflected two papers commissioned by the Commonwealth government: Blocking Content on the Internet: A Technical Perspective and Technical Aspects of Blocking Internet Content.

The latter report, by CSIRO, disappointed the Federal Government - no silver bullets - but is consistent with the concerns about the effectiveness of filters identified by the Electronic Privacy Information Center (EPIC) in its December 1997 report.  

That document demonstrated that many of systems hyped by the US (and Australian) government prevented access to such cesspits as the American Red Cross, the San Diego Zoo, Amnesty International and the Smithsonian Institution.  

It also questioned the credibility of rating services, eg NetShepherd's claim to have rated "97% of the English language sites on the Web". Further reports have been commissioned from CSIRO by the Commonwealth's new NetAlert community awareness body.

Local advocacy group Electronic Frontiers Australia (EFA) offers an Australian perspective on content rating and filtering proposals.

There's a succinct and intelligent discussion of issues and technologies in Geoffrey Nunberg's January 2001 article The Internet Filter Farce. Nunberg is a distinguished science who edited The Future of the Book (Uni of California Press, Berkeley 96), discussed in our publishing guide.



   next part  (censorship offline)