|"We are drowning in information but starved for knowledge." - John Naisbitt, "Megatrends"|
For many researchers who have spent careers studying and attempting to understand the sources and constructs of knowledge, this sudden interest in KM must seem very odd and very superficial. How can you manage something if you don't know what it is? How can you manage the most complex and perplexing thing on earth? How can you manage something as ethereal and temporal as knowledge?
What is knowledge?
Knowledge is obviously not something new but something very old - as old as man. The codification of "knowledge" is also very old. Cave drawings, Aristotelian logic, and language itself are all examples of codified knowledge. But likely more than any other entity, physical or abstract, knowledge is amorphous. It is constantly changing, it is constantly being reborn.
Cognitive scientists have gone to quite a bit of trouble and debate to at least classify knowledge into four categories. Those categories are perception, reasoning, memory and language. Each of these categories relates quite well to the way our brain seems to work. We perceive or sense through sight, sound, touch, taste, and smell. We remember by drawing upon our mental store - some recent events, some distant. We reason through processes which often confound us but are largely based upon drawing analogies to what has worked in the past. And we declare, through verbal and written language (or many other symbol systems), the results of the mixing of perception, memory, and reasoning.
How might we best codify knowledge?
Other researchers, most notably in the area known as intelligent systems, have gone to considerable trouble to create ontologies for representing knowledge. An ontology, in the strict sense, is the study of the nature of existence. As it relates to knowledge management it could be defined as the study of representational systems. Formal ontologies provide a basis for representing objects, functions, relationships, and events. Once these ontologies are established and then instantiated, or filled with specific cases, they can be utilized for later searches, drawing analogies, or making conclusions.
There are many knowledge ontologies. But some of the better know ones include semantic networks, neural networks, predicate calculus, ID3, conceptual dependency diagrams, and CyC-L. These schema's are all quite complex, and require in-depth study. The Handbook of Artificial Intelligence provides a very effective primer in this area1. These ontologies are also the fundamental technology in use in the rising intelligent agent arena.
Ultimately ontologies are tested for their efficacy to deliver systems which exhibit characteristics which humans would label as "intelligent." If an automated system responds to us in the same way we would expect an intelligent human to respond then we can say the system has artificial (man made) intelligence.
Judging a Knowledge Management System
If the purpose of KM is to make it easier for us to draw upon the lessons learned by our predecessors or our own past experiences, then the criteria for judging a "good" KM system then is the same as those characteristics we would associate with an expert or sage. Just such characteristics were developed by researchers in the field of expert systems back in the early 1980s. But they hold just as well today.
The Two Faces - Formal and Informal
Even if we understand knowledge and it's purpose, the challenge we yet face is how to best codify it. What is the best digital representation scheme for knowledge? What formalism and mechanism will provide the most benefit?
To that end there are two general approaches - formal and informal. The formal approach provides a codified ontology such as one of those itemized above. The informal approach allows the contributors to specify their own ontology. These approaches are not necessarily mutually exclusive, although no commercial product exists today which allows for both simultaneously.
As a application example, imagine a video store such as Blockbusters.
We are trying to determine where the films should be placed on the shelves.
Using the formal approach, a detail analysis of each film would be performed.
Pro forma rules would be established for classifying all the films. For
example if, in a film "Blazing Starships", there were more drama scenes
than romantic scenes and the film star was Harrison Ford, we would classify
it as an Adventure film. Using the informal approach, each customer seeing
the film would write down their opinion in a book about the film. Future
customers would read through the customer comment book about "Blazing Starships"
and decide for themselves as to whether the film was worth watching. A
combination system might read through the customer comments, and if the
majority of customers classified it as adventure, it would formalize that
|Internal Structure||codified, follows ontology||loose, user defined, database oriented|
|Basic Features||Q&A, reasoning, relevance||fast store, sort, query|
|Tracking||end point reasoning derivation||user defined threads|
|Display||narrow focus||angular, executive|
|Commercial Products||CyC||Lotus Notes|
Both formal and informal approaches have their place in the organizational arena. It is likely best to start using the informal approach. Creating ontologies is a difficult process and requires a greater level of discipline and effort than providing an "open ended" tool to knowledge workers. Firms such as the "Big 5" accounting firms and multinational organizations have found success in keeping and sharing their workers informal knowledge using tools such as Lotus Notes. As in any system introduction however, managers must be aware of some of the traps that these knowledge collaboration tools may bring. Readers should refer to Neilson2 for further guidance.
1. Barr, A. Cohen, P, and Feigenbaum, E., eds. The Handbook of Artificial Intelligence. vols I - IV. Addison Wesley Publishers. 1982.
2. Neilson, Robert. Collaborative Technologies and Organizational Learning. Idea Group Publishing. 1997.