Copyright:
2001
Statement of
Capability
Original written for Office of Secretary
of Defense (OSD) in 1999
Rewritten June 10, 2001
Overview: Knowledge Management has
been characterized as being, at its core essence, only about two things (1)
truth finding and (2) community building.
These two things are two sides to a single coin. Using this characterization, it is easy to
conjecture that communities actually form due to truths and in response to how
truth is processed within the social fabric.
In fact, communities of interest and community of practice are common
topics at Knowledge
Management (KM) conferences.
At issue, and of considerable interest to
interdisciplinary science, is a clear definition of what knowledge sharing is
and what personal knowledge is. This issue
is not so much about the computer technology and more about the social science.
In many cases, truth is an “invariant” of a social
collective. These invariants are
expressed over and again and can thus be seen in the word frequency
distributions of e-mail, and other text produced by members of the
community. Not brain science, yet. How the brain manages sense making is
relevant; and yet systems like Autonomy, Semio and tacit do not as yet account
for what is know about human
memory, perception and anticipation.
One may speculate that part of the invariant common
to the mental activity of the members of a community is shared knowledge. This is not a crisp and precise statement as
yet. The speculation is, of course, a
reasonable perspective and argues that social science is the key to the successful
deployment of knowledge portals such as Autonomy. In fact, to the extent the speculation, we may conjecture that
the invariants of concept representations, for example derived from natural
language or latent semantic indexing, are a proper scientific representation to
model community sharing of knowledge.
The conjecture is not perfect, in that some issues remain unaccounted
for. However, the conjecture that
shared knowledge can be modeled with text mining techniques is acceptable as a
starting point.
The software systems produced by Autonomy, Tacit
Knowledge Systems and Semio each have an underlying assumption that truth is an
“invariant” of the social collective. The invariant can also be seen in process
models, in statements about lessons learned, and in best practice templates.
Process models, based on structural holonomy, may be used
in the near future to develop the information necessary for modeling knowledge
sharing. Autonomy’s child company (Ncorp) moves us towards the technology
represented by I-RIBs
and Prementia. However, currently we need the human touch
that can be managed using knowledge portals.
From this starting point, we had hoped to
demonstrate an example of knowledge sharing in a functional but virtual
community. This example would have had
a visualization interface and thus the example stood up for review by policy
makers, as an illustrative artifact.
The Core Proposal (1999): We proposed to create a
large-scale example of knowledge sharing within a sub-community at one of the
national labs. This example would have been created through access to two
resources. The first is a specific
amount of time on ARL supercomputers.
The second required resource is an agent based KM system deployed along
with Autonomy, Tacit Knowledge Systems, and Semio’s software. A comparison could be made to Latent
Semantic Indexing – type models of conceptual content of human discourse.
Computer network based agents would be involved in
the automated validation of knowledge artifacts. The architecture would follow
the knowledge technologies standards that reflect distributed and mobile agent
technology, as well as some principles from emergent computing. A specific methodology is envisioned that
would look for the emergence of new knowledge with visualization
techniques. These techniques would
include the well accepted Pathfinder ThemeSpace visualization as well as some
visualization techniques related to concept space production systems, such as
Cognito or Cyc, and some additional theoretical framework based on limiting
distributions of word co-occurrences ( see Figure 1).
Figure 1: Taken from Ralph Abrams’ work on dynamical models of approach-avoidance conflicts.
Our team would deploy a KM
system using a combination from Autonomy, Semio, and/or Tacit Knowledge
Systems. Distributed agent architecture
would be put in place to synthesize profile and knowledge sharing activities
from three separate KM systems.
A model of knowledge use events
iwas proposed as a scientific basis for a visualization methodology based on
theme representations. A Steering
Committee would further develop the model shown in Figure 1. The scientific
grounding for the model includes an extensive paradigm on the evolution of
specific knowledge foci as experienced in an individual human mind, or at a
different level of organization within a community. Using this model, knowledge use events can be represented as an
attractor within a state space defined by vector analysis of text. The two levels can be separately modeled and
the Process
Compartment Hypothesis mathematical models used to examine the dynamics of
emergence. A computer-based
visualization of these events allows the observation of knowledge use events in
communities.
Knowledge scientists have characterized KM to be
about truth-finding and community-building.
Our discussions, in 1999, have identified several specific science
groups within two of the National Labs (Oak Ridge National Lab and Los Alamos)
where it is estimated that community wide knowledge sharing is high and mostly
not occurring on the Internet. These
communities are conducting basic science and thus are very interested in truth
finding. Migrating some of the
presently occurring knowledge sharing activities to a web based system would be
a good idea, and would be an opportunity to conduct some basic research on
Knowledge Asset Management.
A Tasking System Prototype: Tasking is an important
activity within hierarchical organizations such as military organizations. Process models of tasking can be complete
and can be used to track issues and issue resolutions that occur because of the
task order. Moreover, the development
of the task itself has important, and regularly occurring stages. Process models of task
development are available and can be integrated with models of task
fulfillment.
A community of practice is the community where
knowledge use events are to be discovered and represented in a computer
model. We proposed implementing, within three months, a
community specific knowledge vetting system.
The specific “targeted’ community is not organized
in a hierarchical command structure. A
second order “Executive
Information System”, was proposed to manage the deployment process. This second order system would provide a
vetting process to represent events statistically and categorically without any
private information.
An Executive Information System is a tasking system
with feedback. Process models enable
issue tracking and handle the production of outcome metrics. To see how this is done, we make a
distinction between a process model and a sub-process model. Process models are created behind the scenes
to management the evolution of events along a specific sequence. In most cases, the sequences are
predetermined. However, if a difficulty
blocks the predetermined sequence, then a sub-process may be employed to bring
the event sequence back into line.
A methodology creating process models
is needed. Models
allow an adaptation to uncertain events during complex deployments.
So how is the system able to provide adaptation and
process feedback? The answer is that
when issues arise that complicate the fulfillment of the task, then there is a
re-evaluation of the process model. The
re-evaluation is handled using the Prueitt
Voting Procedure.
Complete automation of process sequences is often
impractical. When complications arise,
we have a capacity to reroute the activity evolution into a sub-sequence that
brings information regarding the complication back to the tasker’s
attention. Such rerouting of the
evolution often can be done with a canned, agile or slightly modified
sub-process. This rerouting of
sub-processes is “agile
process re-engineering”.
When an information-gathering phase is completed,
then a decision support system can be deployed to choose sub-processes for
alignment of original task fulfillment, or to stand down the task and issue a
new task.
A task is most often an order to create some change
in the way resources or individuals are deployed. There is specific structure in any specific deployment, thus even
the smallest task can be viewed as incremental re-engineering of the structure
of the organization. This structure can
be enumerated using any one of several methodologies. For example an AS-IS model of deployment captures the structure
of the present situation.
Process
structuring should have a degree of oversight, as well as the means for
authority figures to step in when the intent of the task order is overcome by
external complications. Moreover, AS-IS
and TO-BE specifications can be trended over time to produce various theories
of causation relative to the organization’s long-term behavior. The theory of causation is auxiliary to any
one of the re-engineering process and produces a second order understanding
about the generic issues concurrent with task generation and resolution. This second order system is statistical and
categorical and is validated over a period of time.
Mill’s Logic
can be deployed.
Tasking is most often a function of tacit
knowledge. Process modeling is thus
often less precise and has greater uncertainty than any of us would like to
believe. The distinction between a
process model and a sub-process model allows additional agility to the
automation of task generation and task fulfillment. Moreover, these process models can be used to develop trending
knowledge about characteristic issues that are found to arise in a somewhat
predictable fashion.
Summary:
Knowledge Management is handling, directing, governing,
controlling, coordinating, planning, and organizing agents, components, and
activities participating in the basic knowledge processes (knowledge production
and knowledge integration) of the Knowledge Life Cycle. Knowledge Management manages a complex
process -- its processes and its products.
1)
We
conjecture that shared knowledge can be modeled with text mining techniques and
that knowledge user events will show up as a transition from chaotic type
terminology use to a locally focused terminology use (dynamic model show in
Figure 1 above).
2)
Distinctions
between private and public knowledge can be used to make a corresponding
distinction between the knowledge internal to an organization and the
expression of this knowledge to a group other than the organization.
3)
Knowledge
Management is about complementary processes of "truth
finding" and "community
building". Transitions in
knowledge use events can mark the development and changes that are occurring in
small communities.
4)
Process
models for complementary processes can be used to manage the knowledge sharing
events and the development of knowledge resources.
5)
The
development of new language is a key to enabling proper community building.
6)
Knowledge exists in various modalities, including private and
community; but always involves an interpretation of information.
Such management occurs through a range of activities
including: interpersonal behavior; knowledge processing behavior; and
decision-making behavior.