DAC Home  

Aspects of Knowledge Representation
in Digital Culture
Francisco Ricardo

<back to
schedule
 

This presentation is a study of practices involved in encoding knowledge in digital form. It examines the interface between cultural and formal knowledge, and some of its problems and forms in digital media, especially from the formal standpoint of three classes of knowledge system: machine learning systems, expert systems, and intelligent software agents. The notion of knowledge is itself often too abstract for direct discussion and is invisibly subsumed within concrete systemic or architectural considerations such as document management or hypermedia systems design.

The problem of augmenting knowledge through technology was broached long ago by Doug Engelbart and even before then, by Vannevar Bush. The plan of attack was at first social one: on the premise that knowledge is a collective phenomenon, augmenting it meant supporting collaborative communities of knowledge workers in the course of their work. Here, the system would contain documentary items of day-to-day importance, that is, massive and growing corpora of information. These were the prototypes of the first knowledge management systems.

This approach of providing a technological means by which distributed groups of workers could incorporate information into a growing system accessible to others was the opposite one taken by a later class of knowledge device: the expert system. In this approach, the expertise of a single or select group of domain experts was encoded in a way that a large number of problems in that domain could be solved interactively and without the human expert's presence. This class of knowledge system captured not information but reasoning.

Here are both systems then, one to carry out the convergence of a large body of documents and one to carry out the cognitive specialization of an expert. Vannevar Bush's memex as "a device in which an individual stores all his books, records, and communications, and which is mechanized so that it may be consulted with exceeding speed and flexibility" is a typical mechanism of knowledge convergence. On the other hand, IBM's massively parallel chess playing programs like "Deep Blue" are examples of systems for knowledge specialization. Likewise, there are two opposed dimensions of knowledge practice in the modern social sphere. There is the attempt toward convergence of information and knowledge by codification of these into superstandards like XML. At the same time, the same social sphere is increasingly emphasizing the importance of specialization of the worker into knowledge administrators, knowledge coordinators, programmers of many different types, operators, managers, content designers, technical writers, and so on.

All three types of knowlege engineering architecture involve navigating a search space. The idea of converging spaces like document stores into searchable libraries raises questions about the nature and kind of knowledge one can derive from convergent systems. It is tempting to hypothesize that if we formalize and unify content within a single format and make it efficiently searchable, then we somehow get digital knowledge -- the more content, the more knowledge. This certainly has been the factor of success in powerful specialized systems like Big Blue, which although it is a reasoning system, traverses millions of moves in parallel prior to deciding on a final move. Here quantity has a quality all its own. But is higher quantity-of-knowledge the solution? Some critics of technology have conversely alleged that we already are inundated by too much knowledge or information, and that what is missing is a means or ethical center through which to decide on the right course of action for the world. In formal systems, the convergent solution appears in the form of decision support systems, but nonetheless begs the broader analysis of the role of cultural sign systems in knowledge production and communication ("cultural" here being taken to mean that which transpires with mutually understandable meaning in a closed social system, whether a society, a corporation, or a collaborative).

In a final abstraction of the analysis, fundamental distinctions of class are also discussed, including that between truth -- a first order ontological condition -- and knowledge -- a second order epistemological condition, and that space between truth and knowledge, the phenomenological dimension, in light of leading knowledge representation schemes and sign systems, including intelligent/autonomous agents and expert systems.

 

DAC2001 Home