Armstrong Consulting
1200 Dale Avenue #100
Mountain View, CA 94040


Date: Wed, 12 Sep 2001 19:01:46 PDT


From:   Eric Armstrong
eric.armstrong@eng.sun.com

To:     unrev-II@yahoogroups.com

Subject:   Semantic Technology for Complexity


Alex Shapiro wrote:

...With time, older nodes fade (or become more remote in some other way) and eventually disappear. Nodes can be kept visible, or if invisible then returned to visibility, by linking to them.

Quite reasonable. That's a different kind of prioritizing -- a "fading to oblivion", but it seems quite useful.

Alex Shapiro wrote:

a) In any series of sibling nodes, the highest-rated comes first.

My vision of this one, is that only the most relevant points are actually shown as part of the graph.

I'm still of the opinion that a graphic display mechanism only works for small demos, because the complexity quickly grows too great with respect to the available display area. Given whiteboard-sized LCDs, I *may* be persuaded to change my mind. (I'll have to see, to be sure it works.) But I'm pretty darn certain that graphic displays of complex, interrelated information, simply will not fly with today's display devices.

Alex Shapiro wrote: 


       b) Summary-attempts *replace* the threads they summarize in
           the hierarchy. The previous material is subsumed under the
           summary. That summary may be amended directly, or a
           counter-summary may be offered. In that scenario, a summary
           is always an "alternative" or "idea" that permits other items
           to live in parallel.

Yup. Summaries are very important. A summary can be another type of post that is encouraged by users, and that references all the nodes that it summarizes. Another way to display summaries, is like hints (the way the TG LinkBrowser does it). Basically, users can either be explicitly forced to summarize, or an option could exist for others to summarize for them. 


      c)  Some sort of voting activity takes place, either within the
          system
          or outside of it, and an alternative (aka idea) is promoted to


          the level of "answer". At that point, it goes way up to the top.
          ALL of the questions it answers (since it may well be a solution
          to more than one problem, move UNDER that item, under the
          heading "Why".


           Under  each of those questions, in turn, come all of the alternatives
          that were considered, as well as the reasoning surrounding the
          eventual selection.

I haven't thought about the question-answer structure here, but voting is definitely important. To be honest, I would explore the possibilities of a less structured discussion then one where nodes are labeled as questions/answers. Then again, the question/answer division might be very natural and easy to implement.

From another post, I argue that it is important to allow structures with no typing at all -- the classical outline structure. But it should also be possible to add those types in proactively or retroactively.

Alex Shapiro wrote:

The ultimate goal as I see it is the creation of a "Collaborative Rewritable Document Editor".

We've got code reuse, but not text reuse. So much time is wasted by scientists and journalists all over the place on simply rewriting what has been said before them. Wouldn't it be nice if people were able to settle on an accepted description of a certain issue, and then refer back to it, rather then rewriting the material. This would create symbolism on a higher lever then just words. Paragraphs would come to be reusable tokens. If someone thinks that they could say it better, then they could try, and then people could vote on which version they like. Ok, I have more to say on this issue, but I need time to gather my thoughts. Maybe if someone disagrees, it would help me to form a response.

I think that is close to a good definition of the target. But it needs to carry connotations of "conversation" and "document aggregation", as well. My "HowTo" folders contain dozens of messages with little factoids I've gathered on various subjects. That knowledge base needs to be sharable and searchable.

Interestingly, its contents probably need to be sortable by people using DIFFERENT rating scales. As a beginner, the "10 things I need to know to get started" would be the most important. Later on, the "8 things I keep forgetting how to do" would be the critical information. At an an advanced stage, the "6 things that only experts know about" would want to be sorted at the top. As I progress between those stages, the items I consider valuable would be organically changing, with new and interesting items at the top.

Such ratings are context-dependent. My rating of a particular thing depends on my use for it. There are also "intrinsic ratings", but I'm not sure how to represent both in the same system. To clarify the difference:

  1. The hammer is bad (relative to some ideal hammer standard)

    --An intrinsic rating for a hammer build with a 5lb sledge-head and a pencil-thin handle. It simply ain't gonna work. It's going to break, and be useless, so don't waste your money on it.

  2. The hammer is bad (relative to the task of cutting a board in two).

    --You can use the hammer, but you'll get lots of splinters and jagged edges. A saw is much better for the purpose you have in mind.

Thinking as I write (as usual), it seems clear that a rating is always relative to *something*, and the concept of a rating should probably include a referent that identifies the nature of the rating.

Sincerely,



Eric Armstrong
eric.armstrong@eng.sun.com