Initial Concept

Visual Knowledge Structure

Created 6/10/00 - Last updated 6/11/00 - Jack Harich - Go Back

Status - Possible direction creation. Note humor at end.

Introduction

A tiny wisp of an idea has been fluttering betwixt my ears for awhile. Recently, on long hikes in the Applachian Mountains, it has grown rapidly into an abstraction that has drawn much of my creative imagination. Here's the first written essence of the idea, directly from my May 2000 hiking journal:

Nutshell Vision - A way to transfer bodies of knowledge ten times as fast and accurately.

Narrative - This is currently done with language, emulation and practice. Language is the main way, but has ancient roots to pen and speech that limit it to doing text and sound best. Leap forward by use of modern mechanisms - color, 3D, computers, AI, etc. We envision a small set of "knowledge nuggets" representing the core of any area. There are connected to sub and supersets. Quick, intuitive navigation allows quick learning and reference.

Basically we replace the alphabet with something else as the language foundation. Will still use alphabet to represent discrete knowledge, but not relationships, navigation, emphasis, etc.

Try with very small body of K first. Expect to use "facts".

From my sketch book, we have this first exploration:

This used Tic Tac Toe as a small, doable body of knowledge. The sketch tried to portray the knowledge visually, rather than texturally. Though incomplete, it did pretty well. For example, starting at the word "Game" you can follow the Visual Knowledge Structure (VKS) to the "Name" node, giving the knowledge "Game has Name". Following that to the "TTT" node (short for Tic Tac Toe) one learns that the "Name is Tic Tac Toe", implying the game is named Tic Tac Toe.

Consider Game as the level 1 node or knowledge area root. Of great interest are the level 2 subnodes Name, Rules, Strategy, History and Insights. Notice how these can be reused for any game, allowing one to both record and learn faster. Reuse strikes again! :-)

Some characteristics of Tic Tac Toe are common to other games, such as Chess. The VKS for Chess would link to some of the nodes above, as noted. This is an early glimpse of the importance and complexity of navigation and of the usefulness of "shared" knowledge.

Also from the sketchbook we have these observations:

  1. Huge amount of data appears fast in a body of knowledge.
  2. Try transluscent layers, such as for zooming, additional data, navigational path.
  3. Can brighten layers of choice.
  4. Left click zooms 1 layer, right click unzooms.
  5. Can drag node to see what's around it.
  6. Different colors per layer.
  7. How to navigate from Tic Tac Toe to Chess at low level?
  8. Can "shake" node and it becomes topmost.
  9. Choice of node groups is crucial.

What did we learn here?

  1. A crude VKS can be done with current mechanisms.
  2. A VKS can hold lots of knowledge per square inch, reminding us that "A picture is worth a thousand words."
  3. A VKS can present raw knowledge at least as well as text, if simplicity and unambiguity are goals.
  4. VKSs opens the door to things normal text can't do, such as multiple links per node, structure shaking, flyover detail revelation, color or size for emphais, etc.

Enormous Possibilities

While the above has lots of kinks to work out, the possiblities are entrancing:

1. As I explained to a hiker at Elmer's, who was also an artist, we could summarize and standardize the essence of a body of knowledge easily. This could be applied to beneficial areas like Why Recycle, What An Ecosystem Is, Benefits of Organic Food, etc. One could engineer the content and results of educational thrusts much better, hopefully leading to better results before it's too late. :-( This is a bit utopian, but you get the idea.

2. My website is really just vast reference material. A problem is it's hard to find factual data, hard to see why certain things are true, hard to see the general sweep and high level flow, and hard to separate fact from opinion and explanatory fluff. A good VKS that ran in a browser would fix all that, for my site and many others.

3. A VKS should be excellent for working out a complex problem, and even storing all your research and reference material. (or at least links to it, which are transparent to the VKS reader) When you've been using a VKS for years, it starts to reflect your thoughts and approaches pretty well. Imagine that others could somewhat "enter the mind" of super minds like Bill Joy, James Gosling, Linus, etc. It might take weeks or months to explore the extensive, provocative, humorous, stellar structures such minds have created, but I'd love to take the trip. :-)

4. Groups could enter a "state of flow" more easily.

5. Books or Bodies of Knowledge (BOK) could be published in VKS format electronically. College courses might revolve around a professor's current BOK. The top book categories might become fiction, non-fiction and "fact".

6. Political candidates would publish a BOK on their positions and plans, using an agreed skeleton for all candidates. It is much harder to waffle with the facts a VKS contains. :-)

7. Once use of VKSs became intuitive, they should reveal "knowledge holes" just as deep understanding of chemistry and physics has lead to predictive discoveries.

8. A VKS is concepts connected by relationships, called facts. Since much knowledge is opinion and not fact, or occurs less that 100% of the time, we could put Percentage Probablities (PP) on uncertain relationships. This could be done in a clear, concise, intuitive, unambigious manner. This feature alone would vastly separate VKS from language as we know it. Personally for example, I'd love to eliminate the phrase "up to" and all the hype such phrases are usually found with.

9. Once we have facts and Percentage Probablities, we can take the next step and represent reasoning. One difficult aspect of prose is how it leads to conclusions, often the most important point of written work. With VKS opening the door to clear, tracable, unambigious reasoning, we can more quickly agree or disagree with an author's conclusions.

10. A VKS is just highly organized Declarative Knowledge with a visual renderer, just like a browser renders html. An inference engine, goals and additional VKS data (such as from other VKSs) allows conclusion miners to sift and explore the world's VKSs for incredible results. Once the highly structured data available in many VKSs reaches a certain critical mass, we may see a derived conclusion goldrush. The algorithms for this already exist in expert system and neural net technology, though these have never (?) been applied to large bodies of data not originally created for inference.

11. We will eventually be able to automatically convert millions of web pages to standard VKS format, either from text or underlying XML..

12. It may turn out that behind languages lies a crystalline like structure (or such) that has evaded detection for millenia. It may even be that the brain itself contains a huge, uniform structure that mimics VKS. Or it may turn out that biological mechanisms are incapable of what a hardware/software VKS can do. Hmmmmm, interesting....

13. Knowledge transfer would occur more rapidly. Those who've "learned how to learn" realize that each new field or area is characterized by only a few dozen key concepts. Once you grasp these, the rest is a quick drill down to areas of deeper interest or need. So the new way is you go to a trusted expert's VKS, navigate to the desired knowledge area, scan the topmost facts, and proceed. Wouldn't it be a cool breath of fresh air to learn that way, at mindspeed, instead of wading through books and webpages for the few knowledge nuggets that together compose an area's top layer? As a friend of mine, Derrold Holcomb said once back in the 80's, "I'm reading whole books on building your own sailboat. Lately all I find in some of them is one or two things I really need to know."

14. There are cultural ramifications. We currently teach children two languages - spoken and then written. Introducing a third, VKS, means shuffling the way teaching occurs at home and school quite a bit. Then again, if VKS is that enthralling, it may replace TV with something of substance, (:-) and the problem is more than resolved....

Conclusions

This is a hard problem.

VKS is not html text, images and hyperlinks. We don't know what it is yet, but it's not that.

All in all, I envision VKSs to be vast, pleasurable worlds we live in much of the time, just as we live in books, editors and on line. Rather than pure 3D worlds (which run very slow on today's hardware) one can imagine layers of transluscent images, with side controls such as dials. This is all like the cockpit of a plane, a very studied, optimized and reliable data environment.

We've got to surpass the mouse and keyboard. We need effortless devices such as eye pointers and clickers, head nudgers, lip pushers, eyebrow raisers, brain writers. At first we can do without them all, but they will appear....

Let's get started!!! ;-()

A reasonable first effort would be to take ultra small bodies of knowledge, quickly capture them in a handwritten format, and see if we can get to a point where that's better than text. We can use acetate for layers, colored markers, string, gold stars, folded paper, etc. My guess is proceeding with a software representaiton initially would bog us down, or maybe we'll be surprised and discover we don't need software at all. Once we've explored the problem a bit, we can research. Probably the first major milestone is "Is this a doable problem and solution path?" If no, we redirect, simplify or abandon. If yes, we're in for a bit of fun.... :-)

Breaking New Ground

Much work in this area has already been done. IMHO, initially proceeding without research will be far more productive, because we will not be distracted, swayed and blindsided by other approaches. What's needed here is not lots of work, but a small number of correct concepts organized in a cohesive new direction.

In the small, none or a few of these concepts will be new. In the large, the way they're organized needs to go beyond current work, because we're attempting to create a next generation language. Languages are very different animals from AI knowledge representations, mind maps, semantic networks, data trees, etc. Hopefully this shade of difference has been explained and emphasized enough....

Our definition of language is "that which conveys desired meaning across a large population". Note this doesn't restrict us to mere spoken or written vehicles.

Most engineers get an idea, research it, and then proceed. Why is this not optimal for breaking new ground? As Larry Constantine says on page 80 of Software Development, June 2000 in a dynamite, insightful essay modestly titled "Inventing Software":

"I have had a lifelong habit of trying to work out problems for myself before looking to see what other have done. After I have struggled with a problem on my own and worked out some solution - or at least established a direction - I will then go back and check my thinking aginst the established canon and the received view. Not having immersed myself in prior work, I often find that what seemed obvious to me had escaped many others. Bingo. Another breakthrough. Extensive research and background reading may be a sound formula for writing an academic paper or preparing a thesis, but in the real world it can mire your thinking in the well-worn and muddy ruts already traveled by everyone else."

Oxymoronic Fun

A medium large problem like this may take a small crowd to solve, though it could become a full-time hobby requiring amateur experts. Should we throw a work party, invite a few clever fools, honest politicians and selected volunteers, and hope for awfully good results? No, the second best way is to just have a quick friendly argument in loud whispers, hoping for partial success or at least only a near miss due to bad luck. This is real fantasy, so if we just act naturally and accept all unbiased opinions we're sure to make no deliberate mistakes and avoid a minor disaster.

Then, when we achieve a near optimum solution, we can share it with the military intelligence that so desperately needs a small miracle to bolster its apparent image with deeply superficial knowledge (hmmmm, a double oxymoron) of anything, even vague subjects.

Then again, how about if we rush out a textural version of the visual language? Shucks, that's an unacceptable solution. Sounds terribly nice, but the sweet sorrow of missing such a sure bet will leave us all with a few nights of restless sleep. After all, "If we do not succeed, then we run the risk of failure."

(-: With credit to backpacking.net/oxymoron.html for most source material :-)