Knowledge Matters

Now, consider the notion of knowledge in more detail. A decision maker possesses a storehouse of knowledge, plus abilities to both alter and draw on the contents of that inventory. This characterization holds for all types of decision makers—individuals, groups, and organizations. In the multiparticipant cases, both knowledge and processing abilities are distributed among participants.

Knowledge is extracted on an as-needed basis from the inventory and manipulated to produce solutions for the flow of problems that constitutes a decision manufacturing process. When the inventory is inadequate for solving some problem, outgoing messages are used in an effort to acquire the additional knowledge.

The solution to each problem arising during the manufacturing process is itself a piece of knowledge. In turn, it may be used to find or solve other problems, whose solutions are knowledge allowing still other problems to be solved, and so forth, until the overall problem of producing a decision is solved. Thus, knowledge is the raw material, work-in-process, byproduct, and finished good of decision making.

If a system has and can use a representation of ‘‘something (an object, a procedure,... whatever), then the system itself can also be said to have knowledge, namely, the knowledge embodied in that representation about that thing”. Knowledge is embodied in usable representations, where a representation is a pattern of some kind: symbolic, digital, mental, behavioral, audio, visual, etc. To the extent that we can make use of that representation, it embodies knowledge.

Of particular interest for DSSs are the representations that a computer can use and the knowledge processing ability corresponding to each knowledge representation approaches permitted in its portion of the knowledge storehouse. A DSS cannot process knowledge that it cannot represent. Conversely, a DSS cannot know what is represented by some pattern that it cannot process.

When designing or encountering a particular DSS, we should examine it in terms of the possibilities it presents for representing and processing knowledge—that is, the knowledge-management abilities it has to supplement human cognitive abilities.

Over the years, several computer-based techniques for managing knowledge have been successfully applied to support decision makers, including text/hypertext/document management, database management, data warehousing, solver management spreadsheet analysis, rule management, message management, process management, and so forth. Each of these techniques can represent and process one or more of the three basic types of knowledge important for study of DSSs: descriptive, procedural, and reasoning knowledge.

Knowledge about the state of some world is called descriptive knowledge. It includes descriptions of past, present, future, and hypothetical situations. This knowledge includes data and information. In contrast, procedural knowledge is concerned with step-by-step procedures for accomplishing some task. Reasoning knowledge specifies conclusions that can be drawn when a specified situation exists.

Descriptive, procedural, and reasoning knowledge can be used together within a single DSS to support decision making. For example, a DSS may derive (e.g., from past data) descriptive knowledge (e.g., a forecast) as the solution to a problem by using procedural knowledge indicating how to derive the new knowledge (e.g., how to calculate a forecast from historical observations).

Using reasoning knowledge (e.g., rules) about what procedures are valid under different circumstances, the DSS infers which procedure is appropriate for solving the specific forecasting problem or to infer a valid sequence of existing procedures that, when carried out, would yield a solution.

Aside from knowledge type, knowledge has other attribute dimensions relevant to DSSs. One of these dimensions is knowledge orientation, which holds that a processor’s knowledge can be oriented in the direction of the decision domain, of other related processors with which it interacts, and/or itself. A DSS can thus possess domain knowledge, which is descriptive, procedural, and/or reasoning (DPR) knowledge that allows the DSS to find or solve problems about a domain of interest (e.g., finance).

It can possess relational knowledge, which is DPR knowledge that is the basis of a DSS’s ability to effectively relate to (e.g., interact with) its user and other processors in the course of decision making. A DSS may also have selfknowledge, which is DPR knowledge about what it knows and what it can do. An adaptive DSS is one for which DPR knowledge for any of these three orientations can change by virtue of the DSS’s experiences.