Global Sensemaking

Tools for Dialogue and Deliberation on Wicked Problems

Features and functionality of a global-scalable, collaborative, deliberative, asysnchronous discourse system

Would it be possible to start a pre-specification list of all the features and functions we think would be necessary for a comprehensive sense-making tool? I have studied concepts in effective discourse, including argumentation, group problem solving, discussion visualization, collaboration and a number of tools used to support these. It seems to me that most of the historical approaches (IBIS, Delphi, etc.) have been oriented toward smaller groups. Debate tools, too, seem to be oriented toward small group construction of argument supports. Some aspects, say between argumentation and collaboration seem inherently dialectic (and I haven't been able to see the synthesis).

If there is interest in trying to put in one place all of the things we want to get out of a tool (or interoperable set of tools) then I will be happy to extend this thread with some seeds from my own ConsensUs project to show some of what I think is important.

Jack, David, Ankur, Andy and all? Can we build a functional spec here in Ning? Is this a way to proceed?

George
PS. Has anyone thought about inviting Ward Cunningham (wiki) or other CMC type folks to the party?

Views: 48

Replies to This Discussion

Couple of points, but first, I'd like to remind of the DebateGraph that is entertaining a similar subject that David Price already transcluded in another thread here. In that DebateGraph, I added a few nodes that would serve as part of a response to this thread. Should we not use that DebateGraph?

First point: this thread really relates quite closely to my thesis topic. I have heard horror stories about the ways in which some dissertations were sabotaged by "prior art" found elsewhere. I guess I need to look more closely into such matters, but, at the same time, I have already put the lion's share of my thesis in public in the form of papers, most of which can be found online. In the largest sense, my thesis says something like this: all forms of hypermedia discourse, including blogs, forums, emails, dialogue maps, coherence maps, and so forth, are sources of sensemaking information resource producers and hosts. Sensemaking occurs when one discovers the gestalt entailed in various threads of those many resources. Federation of those resources in a subject-centric manner, not unlike the index in the back of a book--but this one covering the entire Web--is a necessary, but not sufficient component of a sensemaking suite, particularly if sensemaking is to be holistic, and global.

Second point: all human hypermedia discourse activities are going to create an enormous amount of noise. Signal to noise ratio is of great important if we are to be effective sensemakers (response to another of George's threads here). We are more than likely to wrestle with many of the same issues that Wikipedia is already wrestling with, some would argue problematically, but that's another branch of the dialogue map that covers this thread. In short, we will probably, make that absolutely, need a reputation and trust system. Please google the paper "Augmented Social Networks" for more on that. With a reputation and trust methodology in place, the topic map that federates all resources can then provide a one-stop index into the vast and growing sea of sensemaking ideas and results, and it can provide user-controlled filters that allow one to ignore that which has been determined to be less valuable. Subject-centric federation, as I define it, is all inclusive: nothing is censored, though some resources will take on very low R&T metric values, thus effectively making them less visible.

I will offer the argument that the processes I just suggested, an R&T facility coupled with a subject-centric index into our universe of discourse will then provoke serious bloggers and contributors to make the increased effort to create cleaner, more thoughtful, more trustworthy resources. That is to say: we have the ability to improve the level of discourse just by creating the right tools and facilities where people congregate--not unlike the village tree or water well from years past.

It's another discussion, possibly an enormous one, where we launch an R&T dialogue, but for now perhaps it's well enough to think in terms of something like "page rank" for each contribution, augmented by a few other metrics that track the authenticity of references cited and so forth.
I'll need to look at the Debategraph entry. I know David set up a model topic tree (to which I added some commentary) but I haven't been into the tools section yet.

Re: first point. This sounds like a job for data mining! One of my efforts has been to investigate ways of using latent semantic analysis to "cluster" similar semantic nuclei as an attempt to grapple with this notion of gestalt, at least as an aid to the human mind. I can't imagine something like 100 experts in a single subject domain creating volumes of material for the rest of us poor schmoes to work our way through. Then what about thousands of contributors? Global-scale? The view/topic focus mechanism, such as in Debategraph, does help one isolate a portion of a subject for more concentrated analysis. But I take sense-making as including integration, bottom up analysis of discussion points to grasp the big picture, the gestalt. I think we will need a lot of horsepower in that area.

Second point: Somewhat same idea. LSA is a kind of noise filter in that you can set the boundaries of similarity so that non-subject documents are excluded from the clusters. In addition, I advocate a mechanism not unlike short-term/long-term memory traces in neural tissue (the brain) wherein only reinforced topics/issues/commentaries make it into the long-term trace category and anything that is basically ignored (non-active thread) gets eliminated. This has proven to be an effective noise filter for brains. I think it has some potential use in maintaining organization in a humongous knowledge repository such as we anticipate.

As for R&T, the reason I invited Ankur T. is that this is his area of expertise and I have become attuned to the issues from listening to his presentations. I hope he will get more involved, or point us to others who will be.
Great comments. Let me build on them.

In the CALO project at SRI, we explored something called Topic Analysis, based on Latent Dirichlet rather than Latent Semantic analysis. The source code for that engine is included in OpenIRIS. There is a large body of literature as evidenced in this Google search. Agree that it is a data mining project.

Let me offer a counter position: I do not advocate eliminating anything. Instead, consider a metric that is part, perhaps, of the R&T engine suggested earlier, that, over time, performs the same function as decay (forgetting) in any dynamical system. Let me explain first that my thinking is in terms of a topic map, a map of an enormous territory. I also think in terms of cultural federation, where biases of any kind are forbidden in the map itself. There are plenty of biases 'out there' in the societies that engage in dialogues that our map(s) will capture (witness the dialogues here on civility). That is to say, social gestures, including not clicking on particular links, become reflected in some metric that renders an information resource more or less easily discovered. All entities (this is a map, not the stories themselves) are just triples in an enormous data store; keeping them around is less an issue than is leaving them out, in my view.

As I mentioned before, the R&T engine is there to reduce infoglut, but scholars of all stripes should benefit from the availability of each information resource. The R&T engine, driven by the people using the system, determines what is valuable and what is not. I believe that the effect is essentially the same as suggested in the ltm/stm metaphor. Having said that, I am in agreement (some would suggest violent agreement) with the ltm/stm metaphor for all the reasons suggested.

I believe it is possible to craft agent-based platforms, implemented in any number of different ways, that allows mining activities to be conducted in a seti@home-like fashion. Each agent receives, say, a URL to harvest, performs its tasks, and reports back to the base coordination system (e.g. David Gelernter's Linda tuplespace) which then uses web services to update the topic map. This allows distribution of work around the planet. My own thesis project is a prototype of that platform.

Bringing in experts in R&T strikes me as a terrific idea. Many thanks!
Great Start.. Here are some of my thoughts/experiences with reliability and trust issues:

Typically trust enhanced information filtering systems work better in two ways:
a. for new users (we call them cold start users) in recommendation systems. Connecting new users to someone or to information that is largely trustworthy provides an immediate 'safety-net' to novices where it feels safe to contribute or consume information.
b. The reliability of information sources is 'vetted' within the system thereby providing greater resilience to 'spam'. If you know you will be judged for your contribution, you are careful about making a meaningful contribution. (Ha! I hope I'm being meaningful here).

Thus including R & T into a discourse framework is certainly of great value.

Just the other day I was brainstorming on the utility of having star-ratings in various social networking and recommendation systems (think Netflix or Amazon). Even deciding on trust mechanisms that depend on rating schemes (1-5 or 1-10) are controversial but that's another thread.

I like Jack's notion of making trust a function of 'forgetting'. If I recall correctly, its similar to the notion of 'habituation theory' in cognitive science. Interesting indeed.

The other big issues with most trust and reliability metrics are:
a. the treatment of trust as a binary value.
b. the treatment of distrust as a binary value.
c. aggregation of trust.
d. transitive notions of distrust. (is the enemy of my enemy my friend?) This one is a mine field in trust/distrust research.

Hope this adds fuel to the fire!
Thanks for inviting me.
Thanks Ankur. Very meaningful contribution, IMHO. 5 stars on the 1-5 scale :-)

Let me unbundle a few things here. I should have been using the term "knowledge gardening" instead of hypermedia discourse in most of what I said above. Sloppy me. One should not blur the distinction between ordinary Web gestures like writing stories in blogs and that which is coming to be known as hypermedia discourse--that entails those gestures associated with a web of contested ideas. The Web home page for hypermedia discourse is here.

I like the enumeration of big issues related to trust metrics. It seems clear to me that there are several thesis topics floating in the R&T subject alone. For my part, I am trying to fold R&T into those processes associated with the reduction of infoglut, raising the signal to noise ratio. The largest process, in my work, is subject-centric organization of information resources, the migration from document-centric to subject-centric computing.

I think 'habituation theory' is appropriate here; nice contribution. I wasn't thinking quite in such terms, more in line with something Douglas Lenat said in one of his many papers about AM and Eurisko, where the context was his 'Agenda' system. In that work, he quite literally built a complex, adaptive system with feedback and decay. If you have feedback without decay, you get saturation and lose the ability to make appropriate decisions. Let me explain.

An agenda in AM and Eurisko and in my program TSC is a "to do" or task list. Operations 'out there' in the program propose tasks. They lace their tasks with 'reasons' (justifications) which become a kind of 'value metric'. The Agenda, when asked for the 'next task', picks the highest-valued task and returns it. In that cycle or otherwise periodically, the Agenda engine decays values for each remaining task just a little bit. That's the decay part. The feedback part comes when some task completes, gets evaluated, and a reward value is passed along to whatever in the system happens to be dealing with the same object, say, a concept that was just tested for similarity to another concept. If it was found similar, a bit of reward is passed around to anything dealing with that concept, including tasks in the Agenda that are also dealing with that concept. The reward makes the task "more interesting"--a kind of 'focus of attention'. In some sense, I see the Agenda mechanism as a metaphor for sensemaking: we find reasons to get interested in some concept, see someone else do or say something about that concept, then tend to focus even more on it. We need ways to avoid saturation (I think one is called Ritalin ;-) Sleep is good too!

Fuel to the fire? Yup. Thanks for joining!
Please decode "R&T" for me.
Sorry for the delay: R&T == Reputation & Trust
George, I'd like to reframe "a pre-specification list of all the features and functions we think would be necessary for a comprehensive sense-making tool" into a requirements effort. That is, I think we have the cart before the horse, although this may just be my own semantic confusion or just showing my age.

I'd like to create a high-level description of the system (or suite of tools) we are going to produce. You might call it a requirements document. Part of that would be a clear statement of the problem(s) we are trying solve. Another part would be a set of use cases we want to address. Yet another part would describe characteristics of the users of the system and features/functions/structure of the information we need to handle (e.g. interfaces).

Once we have that in hand, we can begin to talk about design in a way in which many people can contribute.

Sounds daunting. Doesn't have to be. We don't have to do it all at once. I'm not a believer in the waterfall process, but I am a believer in having some kind of clearly articulated process that everyone involved understands.

I have the sense that many people here have a good grasp on the problem and various solutions, but I'm certain these are not yet a shared conceptions.

I'd like to find one small problem that we can all agree is part of the overall challenge then drive out requirements, functions, tests, and code to address it. (Also find some real world users to tell us if we've done something helpful.) That would serve to invigorate the group activity ("look, we can do something useful") and inspire us to setup (and shake the bugs out of) a working environment.

To me, a "comprehensive sense-making tool" is way too big a challenge to try to tackle as a unit. Is it possible to factor out a small piece of it and work on that?

Offered in the spirit of agile software development and extreme programming.
Andy, If I were to "reframe" what you just said, I'd say it this way (your mileage may vary)...

Start with a statement of the problem, which we might generally agree has to do with migrating discourse on the Web into a space and methodology that encourages more productive sensemaking, including applications of hypermedia discourse in order to seek solutions to wicked problems.

Hypermedia discourse is already providing us with a set of use cases and scenarios. They exist as the many gestures people can make, for instance, while dialogue mapping. The sensemaking part, where foraging, filtering, and framing the problems, where understanding what the problem is and finding the right questions to ask, is the part where we don't yet have a coherent body of use cases and scenarios. We do have tagging, we do have bookmarking, we do have wiring the Web with coherence relations, but do we have a coherent body of use cases and scenarios that wire all those into sensemaking 'best practices' (or just sensemaking practices)? Perhaps not.

Perhaps we need those before we can articulate the requirements for a sensemaking platform. I say 'platform' since I tend to think in terms of software artifacts, several different types, when federated, as a platform. Actually, there is a bottom's up approach, which is to just ask "what's out there?" and then, when we know that, ask "how can I make what's out there serve sensemaking?". What's out there gives me use cases and scenarios. I then can ask if those fit within some imagined sensemaking/hypermedia discourse framework to serve as a platform. To do that, I still need a vision of what sensemaking means, what the use cases are, and so forth. No free lunch there.

To disambiguate: in my parlance, a platform consists of software artifacts, a framework is a combination of philosophical, methodological, and technological solutions coupled with the social--folks using those toys to save the planet. Those definitions are subject to change as I learn more.

Finding one small problem strikes me as a cool idea, except that our real needs are not small in any sense. In some sense, I am contesting the notion that a "comprehensive sense-making tool is way too big a challenge to try to tackle as a unit." I justify that on the grounds that I'm a) too bull headed to believe otherwise (which doesn't get any points in an R&T metric), and b) convinced that a relatively simple platform is adequate to change the world--Occam's Razor and all that. If I want any R&T points for that assertion, I need to say more. I am saying more over at the wiki, but a really brief sketch goes like this: the platform needs to facilitate the following use cases:
1- Tagging/bookmarking as a means of foraging and reminding
2- Annotating as a means of lifting key ideas out of largish bodies of textual material, much of which is poorly written, incomprehensible to computerized harvesting agents, and as a means of simply making statements of what's important.
3- Connecting as a means of wiring ideas up into a coherent story. Note that the acts of lifting ideas out of the ether and connecting those ideas opens the door to contestable assertions. For instance, just lifting a particular idea out of a story, say, an idea that asserts primacy of some concept, creates a new information resource that can become the focus of attention as a contestable assertion.
4- Dialogues conducted in a structured way (think: Compendium, bCisive, DebateGraph, etc) in order to tease out of people that knowledge that is either explicit or tacit, looking for our best ideas as a means of dealing with contested ideas
5- Subject-centric organization of all of that in order to render it navigable, reveal both disparate and agreeable world views that exist around the same subject, and otherwise help improve the signal to noise ratio.

Scenarios describe all the gestures necessary to "do something" with those use cases.

Again, your mileage might vary, but I like to start with the use cases, put together some scenarios (simple or complex problems to solve), then tease out of that the requirements. What I just said above contains a modest short circuit: most of the tools necessary to satisfy the use cases I just articulated already exist or are very close. The game, at least for me, is to wire them together through Web services and see how it works on a real set of issues.

Just in case it sounds like this forum isn't important to my inquiry, nothing could be further from the truth of that notion. For one thing, I think that more than a few people here and at the wiki are already beginning to sense the need for improvements to the tools we are presently using. For instance, the Deki Wiki provides a marvelous automated hierarchical navigation system, which serves well to illuminate issues related to the use of hierarchies as navigation systems: they make assumptions about how people think. That is, as the number of pages grows, the hierarchy starts to grow as well. When that happens, there are differences in cognitive styles that come into play: a hierarchy that serves the needs of one individual may not at all serve the needs of others. There is no universal ontology that describes the world the way every individual sees that world. We need to find improved ways to render our worlds navigable in ways that can serve everybody. In my view, that's a really tall challenge. It won't be solved by any individual effort; it takes a village. Together, here, we are teasing out real issues related to the use of existing platforms, looking for evidence suggestive of necessary improvements.
Jack: "There is no universal ontology that describes the world the way every individual sees that world."

Actually, I'm coming to believe that there is a universal ontology (and epistemology to boot) that every person does use, but that most people are not aware of. That is what I've decided to call "systemness" (not a word but it should be). People naturally think systemically but over fairly small scales of time and space. That they do so is shown by the way we form concepts and relations. That it is limited in peoples conscious awareness is evidenced by the fact that we have these global issues to begin with!

In fact, I will boldly assert the following hypothesis: Systems thinking is how people make sense of their world!

Systems thinking involves not only an identification process (seeing an entity, boundary, in and out flows, etc., blackbox analysis), but naturally attempting to decompose the system of interest (whitebox analysis as much as possible), and integration into its context, dynamic modeling of the system in its embedding environment.

This natural way to perceive and conceive the world about us is both analytical (reductive) and integrative (holistic) and I believe it is the basis for problem solving in a very general way.

Ergo, I would opt for a platform that supports these kinds of processes through its structured approach to capturing knowledge. We've talked about topic maps, i.e. top-down decomposition of topics into sub-topics, which is exactly what I tried to spec in ConsensUs. I still envision a tree structure (with secondary connections relating sub-topics as needed) with an IBIS-like map attached to each node in the tree. But in addition, I would attach a bottom-up capacity to propose solutions to sub-problems followed by re-integration as we move back up the tree toward large scale solutions. Perhaps wiki-like nodes attached to proposal nodes would be beneficial. As would attachments of models ala Robert Muetzelfeldt's interests.

I submit that the systems approach, embodied in a tool of this kind, would be a powerful means for tackling our wickedest problems. It would provide a "natural" architecture/structure for formulating problems/solutions and with cleverness we could attach any special tools to it that seemed appropriate.

Just my A$0.02
Not bad, and many thanks for being so bold.
I, and maybe others, need to toss this one into IBIS and pound on it for a while. I'd like to see more, with references. Thanks.

More Euros for that one!
Systems science is in the early stages of a comeback from the days after Ludwig von Bertalanffy's "General Systems Theory" lost focus. For years it has languished as a unified subject, letting cybernetics, information theory, dynamical systems theory, etc. take front and center of different stages. But today there are many, like myself, who see a resurgence of a unified subject that can form a framework for other disciplines. I see it now as a theater in the round with different actors facing different audiences, but everybody getting a sense of the whole play.

Some colleagues here at UWT and I are developing a BA/BS in systems science, possibly to launch in 2009. At Portland State U. they have Ph.D. and MS degrees in Systems Science, the first in the country to the best of my knowledge. We are working on a possible curriculum collaboration with them.

In any case, the references are disparate but many. In the meantime, if you haven't already done so, this article at Wikipedia, while still immature, has some reasonable pointers.
http://en.wikipedia.org/wiki/Systems_science

As to the assertion that this is natural (though informal) thinking, I will dig out some psyche references when I get back to campus.
Basically I think I am an early proposer of the systems thinking basis of how we perceive and conceive the world around us. What I am doing is drawing from works like Philip Johnson-Laird's mental models, and other psychology of representation works, to build a theory. I am working on some papers/a book reflecting the arguments. I think the arguments are justified, but not widely appreciated. I'm coming at the subject from a systems science perspective - a systems approach to brain function!

I guess that what I am arguing here is that if correct about the natural systems thinking then a platform that bolsters, and formalizes that thinking process would be most beneficial. I appreciate the multiple perspectives perspective, of course, but at a meta-level I suspect that a formal systems approach would support all of these various perspectives since it is neither just reductive nor holistic but both combined.

RSS

Members

Groups

Videos

  • Add Videos
  • View All

Photos

  • Add Photos
  • View All

Badge

Loading…

© 2017   Created by David Price.   Powered by

Badges  |  Report an Issue  |  Terms of Service