Tools for Semantics

Replies to This Discussion

Permalink Reply by Jack Park on August 16, 2008 at 15:18

Andy, you haven't looked hard enough. NOT! You asked me to tell you that, but it's not true. You ask great questions and offer great observations; the simple fact is that any graph/logic-based implementation of knowledge engineering will only get you so far, and no farther. If humans really did use logic, we might not be having this discussion, possibly because we would all be pretty skilled at sensemaking on the fly, we would be much better anticipatory systems than we appear to be, and the big messes du jour would be different than they are now, possibly more tractable, less wicked. Or not.

We have philosophers since before the famous Greeks who have thought about the very issues you raise; not much has changed, except that we now have computers that let us run simulations of our thinking in ever more powerful ways; computational insight is coming at us faster than the speed of light, and we still do logic and semantic networks, and few people, if any, ever really take the time to figure out what semantic means.

I once read a book, the title of which escapes me, that spoke in terms of Bayesian "logic"; it made sense that the more times one experiences various events, the more one remembers them, links them to other events, and so forth. I'm not a philosopher and never played one on TV, and so I cannot engage in any deep and profound issue mapping that uses terms like "realist", "sophist", "existentialist" and so forth; what I carry as a belief system is a jumble of intuitions based on my own interpretations of everything I experience; I have a hunch it's like that for most people (leaving "most" to be defined by readers).

When one writes an ontology of relations, one is making editorial choices; knowledge engineering is like that. What I want to do is to make the smallest possible set of ontological commitments, and then facilitate further ontological choices to be made by stakeholders who use my tools. Cohere, for instance, allows you to create new relation types, but it gives you a rich and well thought out set to start with; in theory, you should be able to make your assertions from the given set of choices, but if not, press on with your own.

Where things begin to fall apart might be seen to grow out of a hodge-podge of personal choices, some of which might be brilliant contributions, others of which might just be nonsense; that's all part of the game, IMHO. That's why you facilitate online discourse where the nonsense can either be wrestled into sense by the community, or some "usability" or "value" metric grows on the nonsense assertion such that it eventually falls off the radar screen.

I will assert that the field hasn't progressed much. I will further assert that OWL and it's underlying representation scheme RDF don't "compete" with topic maps, for topic maps can and have been created in OWL. I will finally assert that topic maps are just good old fashioned AI frame-based representations of an indexical-relational type; they carry with them one particular discipline, if you can call it that, which entails aggressive maintenance of the co-location principle: one proxy per subject in any given map (one frame for any given concept in a knowledge base). Others, including ontologies of all kinds, are similarly free to adhere to that discipline.

Here's a difference worth studying: ontology integration takes a variety of approaches, some of which are lossy. Merging topic maps simply finds proxies that represent the same subject, no matter how they do so, and merge them; the rest come along for the ride without merging. Subtle, sometime profound, sometimes not so profound differences. I did not just say that ontologies cannot be merged in the same way; they probably can.

Purpose-built. That would describe the general field of knowledge engineering, including topic maps. We build topic maps for representing people, places, and things, just as you would do so with any knowledge base. In my own case, I attempt to reuse representations, to create situations in which shared representations exist. After all, one might create a topic map to engage climatologists, and a different one to engage energy sensemakers; but, the two fields co-exist and co-evolve; they cannot be separated, even though each field has its own ontology; many of those ontological classes are really the same subject, but named differently. Others are intimately related to each other in important ways, some of which we might not yet know about -- unconnected dots, as it were. A primary goal in my work is to facilitate finding and connecting dots in a sense analogous to the early periodic table of the elements which gave indications of un-discovered elements. Can that be pulled off? Watch this space.

• ▶ Reply

Permalink Reply by Andy Streich on August 16, 2008 at 9:17

Regarding the Feigenbaum quote and my paraphrase, Jack said,

"I think I see two phases of GSm. Feel free to offer a correction to the vision I carry in my head: Phase 1—precisely as stated in the quote; Phase II—others, including GSm tool builders, using GSm tools, engaged in the building of a large body of codified knowledge around the issues of our times. Well, that's all I have to say about the quote; it's a beauty—it's just that I see it as half of the story."

I agree with Jacks' two phases and want to drive a couple of points home. Here's phase one again:

Gsm is the computer-facilitated cooperation of some people in the building of a small body of codified knowledge around the development and use of sensemaking tools. The vision is that dozens of sensemaking tool builders and users would cooperate.

I'm really good at obvious things, it's a special talent. ;-)

Obvious here is that getting to phase two means actually doing phase one. I know I'm a steady drum beat on this and risk becoming (or have become) background noise, but if we can create a body of knowledge around what makes a sensemaking tool a sensemaking tool and how to develop them, we can get to phase two sooner and with more gusto. I see phase one as the enabling step. If we can't create some sensemaking knowledge about sensemaking tools we likely are just, to put it crudely, pissing up a rope.

How can we suggest to others they use sensemaking tools in their domains if we aren't using them in our own, if we don't understand and make sense of them? More positively, if we do create a body of knowledge about sensemaking tools, then clearly it will be easier for others to create them going forward.

Remember David, Simon, and Mark K all have suggested and offered the use of Debategraph, Cohere, and Collabatorium to facilitate the work of GSm — and I think we all have Compendium on our local drives. That's four sensemaking tools getting zero uptake by a community of sensemaking experts, tool builders, and advocates to do the collaborative work of Gsm.

Partly I just want to leave it there and see how people respond, but I'm not comfortable with that. I see a very positive outcome and want to express it.

I remember reading in one of Mark's papers, or maybe in Revkin's article about Mark's work, how the Climate Collabatorium could or would allow a user to set up and run a climate simulation with a particular set of parameters. I found that quite inspiring. The idea of an average user kicking off execution of a climate simulation and having the initial parameters and outcome available for all seemed to me a brilliant idea from any number of perspectives.

Then I discovered Debategraph and it's free service of hosting maps of arguments on any sundry topic and, through it's front page, highlighting the more interesting ones. It comes with a particularly useful argument mapping design built-in. Struck me as a good design for several important situations.

Along the way I found Compendium, Cohere, and TruthMapping and thought, "Gosh, there's this wealth of knowledge representation, discussion mapping, and argumentation tools that I'd never seen before." As a commercial software developer it seemed to me to be obvious that weaving these tools together had some payoff for everyone. And this is to say nothing of WordSmyth which, as a lexicon, in some sense serves as the base for everything else.

We've all had similar experiences that lead to the question, "Why doesn't everyone see what I see?"

Clearly you guys don't see what I do or have your reasons for not talking about it / disagreeing with it. But, in the spirit of the Emporer has no clothes, I don't get it. Why not more explicit cooperation/interest in interoperability between tools? Is it that you don't see the benefit to the user? Don't have time to think about it? Or we haven't established a framework for talking about it? (That last seems viable to me.)

Or do you think that I'm just wrong about the potential here? If so, why GSm?

Can anyone shed some light?

• ▶ Reply

Permalink Reply by Robert Parks on August 16, 2008 at 18:38

Here is one response to this discussion. First, I'm not sure we've gotten far enough in this discussion to establish a line of argument with respect to sensemaking tools. So DebateGraph, Compendium and TruthMapping may be a bit premature. My thinking on these issues would benefit from a comparison of several candidate approaches to facilitating interactions:
a. open source and other software development models. What procedures from open source interactions would be good to adopt?
b. expert system development using repertory grid for elicitation and socio grid for comparison. Could we use these tools to make our views more explicit and compare them?
c. tools for ontology development, comparison and merging. Could/should we be developing a structured vocabulary to facilitate and target the discussion?
d. tools to simulate and facilitate different communication processes, rules, procedures, etc
e. Better tools to better facilitate this kind of free flowing discussion.

Several people have mentioned different blog/wiki/tagging/discussion tools, and their features. And some tools are being used - such as Diigo and the wiki. But we haven't decided on a method of aggregating/storing/organizing this information. I think we're in the discussion phase, but perhaps the broad agenda doesn't provide a mechanism for creating sub agendas, reporting processes, etc. We have software tools to facilitate direct interaction - such as scheduling tools and calendars. But I'm not sure if there are any convenient software tools for facilitating development and implementation of an online discussion agenda. Am I missing something?
Bob

• ▶ Reply

Permalink Reply by Andy Streich on August 20, 2008 at 3:56

A sincere "thank you" for redirecting my energy, Bob.

"I think we're in the discussion phase, .... I'm not sure if there are any convenient software tools for facilitating development and implementation of an online discussion agenda."

In some sense, this is the very foundation of making sensemaking tools useful on the web. To me your phrase "convenient software tools for facilitating development and implementation of an online discussion agenda" (I'm not aware of any) is a key part of the design of any significant step forward in the utility of sensemaking tools. Current discussion-oriented tools show great progress in the design, presentation, and editing capabilities of nodes (ideas/issues/positions) and connections (relationships). One gap to fill is in goal oriented behavior. We want to achieve some end (e.g. a decision or the creation of a consensus document for a given topic). How do we facilitate that? What set of behaviors are needed? What sort of software aid might guide the process?

We can use our own group, GSm, and its expressed goals as a case in point. What's missing from our environment that might make us more effective as a group working to evolve better sensemaking tools and better usage/integration/federation of existing ones?

Your "a" to "e" list of candidate approaches is a good set to mine.

"e. Better tools to better facilitate this kind of free flowing discussion."

To me this is the big gap. Tools exist for a host of sensemaking activities including foraging, annotating, social bookmarking, tagging, even real-time face-to-face discussion mapping, etc., but really nothing that strikes me as particularly effective for facilitating and "capturing" a free flowing online discussion. Seems to me such a tool must incorporate at least your points "c" and "d" above as well as Jack's TopicSpaces. I wonder what the next best step toward such a tool might look like and note that Ning has an open API (which I've never explored).

"b. expert system development using repertory grid for elicitation and socio grid for comparison. Could we use these tools to make our views more explicit and compare them?"

A great idea for multiple reasons the first of which is exactly what you state. Beyond that these are tools that are well-defined, useful for discovering and disambiguating topics/subjects, and can be incorporated into a GSm tool suite for that purpose. (A couple days ago I re-read some of the PCP papers and finally "got" -- to some extent -- the theory behind the tools.)

• ▶ Reply

Permalink Reply by Scott on August 19, 2008 at 3:23

An answer to resolving intelligent arguments within a group framework is hidden in the complexities of Benjamin Franklin's Junto. What I originally glossed over as being a simple group process, now I conclude is an anomaly which has rarely been duplicated in history.

Egocentricities make it difficult to form a Junto and harder to maintain over a long period of time.

There are many unknowns of what contributed to the longevity and productivity for Franklin's Junto.

Some guesses of what led to Franklin's success:

1. Small group with little turn around or additions. Less time understanding the dynamics and complexities of new personalities. Respect was easier to maintain.
2. Originally, respect was anticipated as being key, but lately I've determined that anger management and a sophisticated communication process were more important.
3. The depth and mutual respect of member relationships prevented anger from becoming rooted.
4. Benjamin Franklin was in his early twenties. I'm assuming the remaining members were below 30. Younger people are less hardened, inquisitive, and open to alternative points of view. Regular meetings at this age could have contributed to a rare indepth respect.
5. The undivided attention given to a single weekly presenter provided a continual openness for each member.

• ▶ Reply

Permalink Reply by Andy Streich on August 19, 2008 at 20:23

Scott,

I would add to your fine list:

6. Ben and cohorts weren't distracted by TV, text messaging, Internet, etc. (Put negatively: what else did they have to do?) Thoughtful dialog was (grossly) analogous to American Idol (just guessing, never seen AI).

7. They were face-to-face when the Junto met. I think this is the really telling point.

8. Life was so very different in the 1700's. Life expectancy about half of today's (Ben himself being an outlier). Women relegated to secondary roles and too frequently dying in childbirth. Most people were farmers. Urban to rural ration was small. Travel of any sort arduous and communication extremely by our standards. Small minority had access to books and anything other than very basic educational resources. Etc.

So the Junto offered much that was just not available in any other form in that era.

• ▶ Reply

• ‹ Previous
• 1
• 2
• 3
• Next ›
• Page