KLMScalingOrganization

Scaling Organization of Content to Pace Decentralized Content Generation

Decentralized content management enables capturing and presenting vast amounts of knowledge, to an extent that can impede, rather than promote, the information discovery process within the captured info. This needs taming.

Centralized mechanisms like automated indexing and search engines provide means to jump into vast collections, but do not help for getting oriented within them - for finding your way around the neighborhood once you're there. Living with only links and searches is like living in a future world where every room is connected to every other by transporters (beam me up scotty). There is little or no cue to immediate context - no neighborhoods, no landscape. It is by answers in context that people get familar with a subject, not by collections of disjointed answers, alone.

Automatic inference of organization based on content (eg, google's iterative topologic sorting of cross-references) and editorial oversight can both help address the problem, but both generally lack the immediacy and insight of those most intimately concerned with the information - the content authors and the visitors consuming their products. There are many opportunities to introduce low- or no-burden measures into the content development and discovery processes, keeping pace with the development and consumption processes by tapping into the decentralized scaling that they involve.

I propose to incorporate provisions for such implicit mechanisms, in addition to reasonable automatic ones, to promote cohesive and comprehensible content organization that aids both the site visitor and the content developer in their intrinsic collaboration about the information the information conveys. Ultimately, the goal is to cultivate connecting disparate answers about different aspects of a subject, collected by multiple authors, into coherent stories about the subject, helping to resolve the underlying story that connects the answers: "turning answers into stories"...

We follow the following principles to keep our self-organization mechanisms managable:

• Easy to use - easy to add stuff, and easy to put it "in the right place".
• Self-regulating - Ability to delegate discretion about regulation of content development and authority. Ie, discretion over delegation of delegation authority...
• Progressive - incremental development, convergent/non-chaotic feedback, ability to pick and choose parts to deploy.
• Explicit - obvious and overt inference of feedback - non-magical computation, and direct, non-invasive information collection.

Here are some prospective primary avenues to focus on.

Content Development Process

• Low impedence, high flexibility, high functionality authoring, with high-discretion delegation of control.
• Maintain page associations

  Maintain page associations in an "organization" resource - and retain the information based on process cues like generation of new pages from old. This kind of info can be adjusted after the fact, but requires no intervention on the document authors to determine "regional" relationships like parent/offspring, useful for things like:

  . Meaningful and comprehensive table-of-contents for the collections

  . New offspring obtain meaningful defaults for characteristic properties like security policies, notification interest, etc, according to the settings of the originating parents

  . Offspring have inherent "next"/"previous" according to their sequence in their parent document's text

  . etc

• Community-refined/extensible classification system:

  Where classification topics and refinements are generated by community member assignments of classifications to documents, with popular choices having greater prominence.

  Topical classifications:

  1. Which people use to identify submissions (of their own, and of others)
  2. Choices at any level are sorted by frequency of use
  3. People can extend at any point with new choices - understanding that their unique choices will be low prominence unless others seek them out (at the bottom of the lists), or independently originate them.

  (To reduce browser round trips, we may want to do some kind of some kind of pre-fetching outline-navigation system. There's a free one called "joust" - http://www.ivanpeters.com - it's a javascript mechanism for managing and navigating outlines. If we have resources, we could probably do something better tailored for this very specific appliction, minimizing use of javascript as much as possible.)

• Automatic full-text and classification meta-data indexing - Zope catalog type stuff
• Sophisticated inter-relationships graphing

  . Google-like iterative topological sorts on cross-references, which discern more central and more peripheral pages

  . IBM "graphing the web" techniques to identify macroscopic structure: http://www.almaden.ibm.com/cs/k53/www9.final

Content Discovery Process

• Monitoring of change monitoring

  Enable community members to register for change notifications to pages, and present statistics about what's being monitored so community members can tell where the attention is going, where the interest currently is, as it changes.

  The change notification registrations, themselves, can offer the option to propagate to whatever depth the member wishes along document lineage lines (see "Maintain page relationships", above). Shallow monitoring means concern with the higher-level "executive summary" perspective, not with nitty-gritty details deeper within the offspring hierarchy. Electing for "deeper" monitoring means getting the geeks perspective - concern for all the details.

• Monitor favorites voting - "buzz"

  (I've got internal notes about this at http://serenade:7290/Artifacts/zwiki/CommunityBuzz - not yet for external consumption. Some overview:

  "Buzz" is a community-driven measure for identifying and promoting attention to items of particular interest. It is driven by simple community-member votes, and provides a basic mechanism for sorting community contributed content according to expressed community interest, based on collected community-member ratings. It works by aggregating optional +/0/- votes on items to assign relative-interest values. A primary requirement, for scaling, is zero administative intervention - it should be based entirely on the aggregate of votes from the community members.

  Does stone society stuff have any bearing? http://home.san.rr.com/merel/ss.html

• Using query-satisfaction feedback to tune searches

  There's probably literature on this - it may be getting into the heavy magic realm, though.

• It may be interesting to collect and collate traversal patterns, usage patterns - like the previous item, if its worth doing, we can probably find literature about it.