Jack Park
jackpark@thinkalong.com


Date: Wed, 11 Apr 2001 18:25:56 -0400


Mr. Rod Welch
rowelch@attglobal.net
The Welch Company
440 Davis Court #1602
San Francisco, CA 94111 2496

Subject:   Peer Review on KM Theory in POIMS


[Responding to your letter on March 29, 2001 requesting peer review on additions to POIMS that explain the KM movement.... ]

Well Rod, I just spent some time rereading Poims. Most of it is quite lucid and repetitive, but worth reading.

You broke the KM jockeys into two groups, those that talk and those that do otherwise. You then marry the talkers with email and proceed to show 4 reasons why that is bad.

Your first point, that there exists (implied: as far as you know) no evidence that AI and web pages can support (your words) an effective architecture for human intelligence. Gads, that's a huge one to hoist aboard.

I suppose that elsewhere in your verbiage, there exist words that properly lay out an *effective architecture that supports human intelligence* I have not seen anything that I can explicitly identify as an architecture that suits that problem space, whatever that problem space is.

What is human intelligence? I gather that we are in violent agreement that human intelligence has a lot to do with "doing the right thing" and that all of the POIMS technology is about helping the human be ready to do just that. You go to great pains to imply (show without proper evidence) that nothing else but POIMS is suited to this task.

I tend to draw some lines of distinction that I am not sure that you do. For instance, email does help me to *do the right thing* on occasion, and sometimes not.

From my scattered readings of your various entries in your online world, I would have to conclude that, on occasion, you are *doing the right thing*, but not always. Case in point: your four arguments against email are poorly structured, and maybe wrong. But that's just my judgement, clearly not yours.

And, that points out one of the problems I have harped on for a long time now with your rendition of an SDS: it's just YOUR rendition, not mine and not anybody else's.

It's just an enormous information space filled with records of events that have crossed your path, and a collection of your musings on those events. Powerful stuff, that, but there is no room for realtime discussion. Rather, discussion comes in the form of clippings from emails or phone conversations (which, on occasion, contain improper or wrong *interpretations* of what you heard), followed by lots more stuff that reads like a muse.

Rod, I strongly believe that what you are doing is of enormous value. But, it serves collaboration only to the extent that you are able to behave as a mastermind, remembering every little thing anybody ever wrote that you came across. That's useful. Sometimes, your musings are not (at least to me) all that useful. Sometimes, for me, they are wrong. And, on those occasions, it may not be personal opinion, rather actual knowledge of facts that greatly differ from those you reflect and react to.

In short, for many of the reasons you say that email is bad, so might be SDS.

Now, that's just my set of interpretations. Your mileage may vary. Do not perform these inferences at home; your health may be at risk.

From my model of collaboration (which, IMHO, is a greatly different behavior than minding the events with SDS), what would be needed is that the muse space of SDS be turned over to a collaboratory, one in which the musings are output to an online issue-based discussion seeking a consensus interpretation of the events in question.

As I have said before, such a collaboratory will need the capability of maintaining a kind of truth maintenance system, one that is capable of spotting such faux pas as self-contradictions, referents not *on the same page* and so forth. Such a system would, ultimately, probably need the ability to go outside the present universe of discourse and bring in other evidence by itself. The system should probably also be capable of weighing the information content of various statements, judging whether they represent an addition to the conversation, or just noise. POIMS, as it appears to be implemented in SDS, does not give evidence that it is capable of doing any of these things.

How might, say, a re implementation of the present SDS move in this direction? My hunch is that SDS should be re implemented to behave just as it presently does, but with the addition that it can submit its musings to an IBIS discussion list and acquire responses for further musing, which, of course, would be presented to the discussion. Simply make SDS a node in a discussion system and you're awfully close to collaboration as I presently see it. The human should stay precisely in the loop you presently implement, but that loop needs to be tweaked.

Cheers,


Jack Park
jackpark@thinkalong.com