CONTACTS 

SUBJECTS
Viability DKR
Bumbling, Mistakes, Meaning Drift
Risk DKR Need Reduncy of PC Avoid Impact of Bumbling
Reliability Redundancy Bumbling
Imploding Critical Mass Mistakes
Risk DKR Failure War Stories

1008 -    ..
1009 - Summary/Objective
1010 -
101001 - Follow up ref SDS 4 0001, ref SDS 2 0001.
101002 -
101003 - Eric Armstrong relates considerations by the Colloquium of problems
101004 - that have and can arise in developing a useful knowledge management
101005 - system, responding to query by Jon Winters.
101006 -
101007 -
101008 -
101009 -
1011 -
1012 -
1013 - Progress
1014 -
101401 -  ..
101402 - DKR Opportunities Balanced by Risks of Bumbling
101403 -
101404 - Follow up ref SDS 4 4928.
101405 -
101406 - Received letter from Eric Armstrong concurring with Jon Winters
101407 - letter, ref DRP 2 0001, received on 000307, urging that the
101408 - Colloquium track "horror stories" about what can go wrong.  Eric cites
101409 - a presentation by a representative from Citigroup who spoke in Session
101410 - 9, and mentioned Murphyisms to be alert for. ref DRT 1 0001
101411 - ..
101412 - Eric does not mention any problems the Colloquium is considering
101413 - in its design, nor does he cite a risk management effort to identify
101414 - potential problems that the design will accomodate arising from new
101415 - realities of a more complex environment, cited on 980307. ref SDS 4
101416 - 3866
101417 -
101418 - In particular the following issues are not discussed...
101419 -
101420 -   1.  How will DKR get useful content when CEOs, managers and
101421 -       engineers don't want to do the hard work required, reviewed on
101422 -       000307. ref SDS 4 4073
101423 -
101424 -   2.  How will DKR expanded knowledge capability avoid boggling the
101425 -       mind. ref SDS 4 4928
101426 -       ..
101427 -   3.  How will DKR avoid continual bumbling that slows everybody
101428 -       down. ref SDS 4 6794
101429 -
101430 -        [On 000317 submited letter to Colloquium suggesting need to
101431 -        develop people with skills to perform knowledge management.
101432 -        ref SDS 6 0005
101433 -
101434 -        [On 000407 listed agenda issue for project meetings.
101435 -        ref SDS 7 3961
101436 -
101437 -
101438 -
101439 -
101440 -
101441 -
1015 -

SUBJECTS
Subject Indexing
Classification
Context Management
Repository of Experience
Subject Indexing is Complex, Others Have Failed
People Need People to Help Identify Subjects
Organic Subject Structure
Subjects Need Expert Help with Searches
Search Procecures, 000309
Friendly Design DKR Risk of Redundant Questions, 000309
Use Case Method Define Requirement, 000324
Ease of Use Reduces Productivity, Jeff Miller, Eugene Kim

2815 -
281501 -  ..
281502 - Subject Searches Rigid, Inflexible Solved by Friendly Design
281503 - People Like to Ask for Help and Have a Human Mind Interpret
281504 -
281505 - Follow up ref SDS 4 2958.
281506 -
281507 - Eric advises the issue Jon Winters cites on 000307 where people ask a
281508 - question that has already been answered, ref SDS 4 2958, is very,
281509 - very important. ref DRT 1 6510
281510 -
281511 -    ..
281512 -   Use Case:
281513 -
281514 -       Person asks a question that has already been answered.
281515 -
281516 -       System Action:
281517 -
281518 -       1.  It is important that the question is *not* entered into the
281519 -           resository until it has been refined sufficiently to be a
281520 -           unique question.(Or it should be possible to delete it
281521 -           later, as well as revise it.), ref DRT 1 0800
281522 -
281523 -           Question
281524 -
281525 -              How is this important objective accomplished?
281526 -              ..
281527 -              Seems to require interface capability, per review
281528 -              on 00304 for JPS phone system. ref SDS 1 3066
281529 -
281530 -       2.  Ideally, the system will be "friendly" enough for people to
281531 -           find things. Most NewsGroups aren't that friendly, because
281532 -           of the large volume of frequently redudant material (see
281533 -           point #1) and the inability to group things together. So the
281534 -           ability to "grow" the organization of the data by adding
281535 -           categories and moving or copying items to them is paramount.
281536 -           ref DRT 1 1647
281537 -
281538 -              Question
281539 -
281540 -              What is the source of "redundant" questions?
281541 -              ..
281542 -              A contributing factor may be defective instructions
281543 -              by vendores who are too busy working on the next project
281544 -              to enable users to apply the last product cycle.
281545 -          ..
281546 -      3.  Even with the best system, a new person will frequently
281547 -          ask a question "in the wrong way" -- using the wrong
281548 -          terminology, etc. That makes automated detection a difficult
281549 -          task.  What's needed is a way that makes it tres simple for
281550 -          another user of the system to respond "answered in X", where
281551 -          X is a link to the appropriate node. That answer might have
281552 -          the effect of adding the question as an "alternative form" to
281553 -          the original, or simply prevent it from being added to the
281554 -          database, unless and until it is redefined and resubmitted as
281555 -          a new question that no one responds to in that way.
281556 -          ref DRT 1 3680
281557 -
281558 -
281559 -   Updating Information
281560 -       ..
281561 -       Eric cites a case study from personal experience.  He sent
281562 -       out a query asking for good optometrists? ref DRT 1 1428
281563 -
281564 -       10 replies were received.
281565 -       ..
281566 -       One reply included 10 replies somone received when he
281567 -       asked the same question a few months earlier.
281568 -
281569 -       There was a 30% overlap between the two lists. Three people were
281570 -       kind enough to repeat the same recommendation they had given
281571 -       earlier. The other 7 were new -- from new people in the
281572 -       organization or because they had come across new information.
281573 -
281574 -       Interesting questions:
281575 -
281576 -         a.  How are questions updated with new answers? It is repeated
281577 -             queries that spark new additions.  What should happen,
281578 -             ideally, is that new answers are linked to the original
281579 -             question, rather than to the duplicate (which should not
281580 -             be entered). Ideally, duplicates are eliminated there, as
281581 -             well. But again, that needs to be a human process, since
281582 -             Jeffrey Archer in one response might be J. Archer in
281583 -             another, and that might be hard to distinquish from his
281584 -             brother, K. Archer, in a third. ref DRT 1 2730
281585 -             ..
281586 -         b.  How are answers updated? If J. Archer changes his
281587 -             address from Cupertino to Pleasanton, that's important
281588 -             information, and the correction should replace the
281589 -             original. ref DRT 1 3774
281590 -
281591 -         c.  How are invalid and out of date responses removed? If J.
281592 -             Archer retired, how/when is that response removed from the
281593 -             repository? ref DRT 1 5418
281594 -
281595 -
281596 -
281597 -
281598 -
281599 -
281600 -
281601 -
281602 -