Knowledge is Associative
While information is eminently descriptive and can refer to the past, present, and future, knowledge is highly associative. That is, knowledge allows us to "associate" different world states and respective mental representations. These representations are typically linked to, or described by, pieces of information (i.e., knowledge allows us to link different pieces of information and make decisions based on associations). The associative aspect of knowledge can be divided into two types, namely correlational and causal, which are, in turn, only two types of what is referred to by Weick and Bougon (1986, p. 104) as "cognitive archetypes."
Correlational knowledge usually connects two or more pieces of information that describe events or situations that have happened, are happening, or will happen at the same time. Causal knowledge connects pieces of information that describe the state of the world at different times. For example, consider the associative knowledge represented in the following decision rule: "If John has a fever and is sneezing, then John is likely to have a cold." The knowledge embodied in this decision rule is of the correlational type because it affirms that someone who has a fever and sneezing is, in fact, displaying typical cold symptoms (i.e., "fever," "sneezing," and "cold") typically happen at the same time.
Another example of a different type of knowledge is provided by the rule, "If John smokes a lot, then he will probably die from lung cancer." This decision rule expresses causal knowledge. As such, the rule connects two events that take place at different times - John smokes a lot in the present, and John may dye of lung cancer in the future. Dennett (1991, p. 144) refers to causal knowledge when he claims that:
The brain's task is to guide the body it controls through a world of shifting conditions and sudden surprises, so it must gather information from that world and use [it] swiftly to "produce future" - to extract anticipations in order to stay one step ahead of disaster [original emphasis].
Knowledge drives the flow of myriad decisions that have to be made even in the simplest organizational processes. Steel plants, for example, rely on process teams to load and operate smelters. Consider the predictive knowledge expressed in the rule, "If the smelter is set at a temperature of 3,000 degrees Celsius, then a 1-ton load of steel will be smelted in 43 minutes." This is one of the pieces of knowledge that allows a smelter operator to predict that a batch of solid steel weighing about 1 ton will be in liquid form approximately 43 minutes after it is loaded into the smelter, if the smelter is set properly. This prediction allows the smelter operator to program a stop in the smelting process at the right time and let the liquid steel flow out of the smelter, which saves energy and, at the same time, prevents the steel from overcooking.
For teamwork to yield effective and efficient outcomes, those who perform activities in a process must share predictive knowledge. In the example, those who use the steel in liquid form for shaping steel parts should ideally hold at least part of the knowledge held by the smelter operator. If they know of the "43 minute rule," they can also predict that a batch of steel will be ready 43 minutes from the time it is loaded in solid form and have their own equipment prepared at the right time to work on the liquid steel.
In general, business knowledge is inextricably linked with decision making (Olson and Courtney, 1992; Holsapple and Whinston, 1996), perhaps because one of the best ways of assessing the actual value of knowledge is through the assessment of the outcomes of decisions made based on it. Holsapple and Whinston (1996, p. 6) talk of the importance of knowledge for decision making:
For centuries, managers have used the knowledge available to them to make decisions [original emphasis] shaping the world in which they lived. The impacts of managers' decisions have ranged from those affecting the world in some small or fleeting way to those of global and lasting proportions. Over the centuries, the number of decisions being made per time period has tended to increase. The complexity of decision activities has grown. The amount of knowledge used in making decisions has exploded. There is no sign that these trends are about to stop. If anything, they appear to be accelerating.
Knowledge has been distinguished from information. It is also linked with decision making in different fields of research and academic disciplines. In the field of artificial intelligence, for example, information has been typically represented through "facts." Knowledge, on the other hand, has been expressed by means of a number of different representations, such as semantic networks, frames, scripts, neural networks, and production rules; the latter is the most common in practical knowledge-based computer systems (Callatay, 1986; Holyoak, 1991; Olson and Courtney, 1992). Production rules are conditional statements in if-then form, like the ones used to exemplify knowledge in this section.
In the fields of psychology and social cognition, knowledge has been expressed through schemas (Lord and Foti, 1986) and cognitive maps (Weick and Bougon, 1986). These are, in turn, seen as guiding individual and group behavior and using, as input, environmental stimuli obtained through the senses. The concept of schema was developed as a reaction to studies of memory pioneered by Ebbingaus, who made use of arbitrary materials and sensorial stimuli to determine factors that influence the formation of memory and recall of information (Gardner, 1985). The development of the concept of schema is credited to Bartlett (1932), who used an Indian folktale called "The War of the Ghosts" to show that existing mental structures strongly influenced memory formation and recall. Such existing mental structures, which were used by Bartlett's study subjects to process information coming from the tale, were called schemas. Essentially, Bartlett has shown that individuals possessing different schemas would interpret the tale, which is filled with strange gaps and bizarre causal sequences, in substantially different ways.
In biology (more particularly in neurology) knowledge is typically associated with long-term, nerve-based memory structures that mainly process information (Pinker, 1997). Information is seen as usually associated with short-term neural connections that appear to "vanish" from conscious memory after a while. For example, the knowledge of how to operate a telephone is stored in long-term memory structures, whereas the information represented by a phone number is stored in short-term memory structures.
The Value of Knowledge
Knowledge is usually much more expensive to produce than information. For
example, information in the form of mutual fund indicators (e.g., weekly
earnings, monthly price fluctuation, etc.) is produced by means of simple
calculations performed on data about share prices and their fluctuation over a
time period. The knowledge of how mutual fund indicators fluctuate, however,
requires years of analysis of information built up over time. This
analysis of information leads to the development of knowledge that allows an
expert investor to select the best mutual funds according to the configuration
of the economy. This leads us to the question: How is knowledge produced?
Comparative studies of experts and nonexperts suggest that expertise is
usually acquired through an inductive process in which generalizations are
made based on the frequency with which a certain piece of information occurs.
These generalizations are the basis for the construction of knowledge (Camerer
and Johnson, 1991).
A different and less common method used to generate knowledge is deduction,
whereby hidden knowledge is produced based on existing knowledge taken through
a set of logical steps (Teichman and Evans, 1995). This method has been used
in the development of a large body of knowledge in the form of "theorems,"
particularly in the fields of mathematics and theoretical physics (Hawking,
An example of knowledge-building through induction is that undergone by novice investors
in the stock market. The observation that shares of a small number of
companies in high technology industries have risen 10 percentage points
above the Standard & Poor's 500 Average Index during a period of 6 months
may prompt novice investors to put all of their money into these shares.
However, a professional investor with 10 years experience as a broker
in the stock market knows that a 6-month observation period is not long enough
to support such a risky decision and opts for a more diversified portfolio. In
cases such as these, a novice will probably sell, based on the same pattern
used when buying, and will eventually lose money. These decisions were based
on inferences based on a time span that is too short, leading the novice
investor to buy shares that are overvalued and sell these shares when they are
undervalued. According to Boroson (1997), most nonprofessional investors
follow this recipe, which, in most cases, leads to disastrous
The example above illustrates a key finding from research on cognitive
psychology - people usually tend to infer knowledge based on the observation
of a small number of events, that is, on limited information (Feldman, 1986).
Moreover, once knowledge structures are developed, changing these structures
can become more difficult than developing them from scratch (Woofford, 1994).
A conversation that one of us (Ned Kock) recently had with a university
colleague illustrates these cognitive biases. The colleague had gone to two
different agencies of the New Jersey Motor Vehicle Services (MVS) where he met
employees who lacked sympathy and friendliness. He also had gone to a similar
agency in Pennsylvania, whose employees he found to be very nice. Later,
during a chat with friends, he said:
"... All MVS employees in New Jersey are very grumpy, difficult to deal
with ... The state of Pennsylvania is much better in that respect ..."
He had just made a gross generalization, given the small sample of MVS
agencies visited - two in New Jersey and only one in Pennsylvania. Although he
agreed this was a generalization, he was never-theless adamant that he would
never go to a New Jersey MVS agency again, unless it was absolutely necessary.
If this was the case, he said he would ask a less "touchy" person to go - his
The development of theories of knowledge (or epistemologies) and scientific
methods of inquiry has been motivated by a need to overcome cognitive biases
as illustrated above. This has been one of the main common goals of such
thinkers as Aristotle, René Descartes, Gottlob Frege, Bertrand Russell, Karl
Popper, and Thomas Kuhn. Epistemologies and scientific methods have provided a
basis for the conduct of research in general and, in consequence, for
technological advances that have shaped organizations and society. Every year
hundreds of billions of dollars are invested in research with the ultimate
goal of generating highly reliable and valid knowledge. And the market value
of organizations is increasingly assessed based on the amount of knowledge
that they possess rather than on their material assets base (Davidow and
Malone, 1992; Toffler, 1991).
Paul Strassmann, a former information technology executive at organizations
such as Xerox, Kraft Foods, and the U. S. Department of Defense, suggests that
variations in the perceptions of organizational knowledge account for the
growing trend toward overvaluing or undervaluing common stocks in the share
market. According to Strassmann, the perception that a stock is overvalued
stems from the failure of current accounting systems to account for the
knowledge assets of organizations, and he presents an impressive array of data
to support this idea. Abbott Laboratories is one of the companies he used to
illustrate this point.
Over a period of 7 years from 1987 to 1994, the ratio between Abbott's
market value (defined by stock price), and its equity has swung from five up
to nearly eight and back down to about seven. However, the ratio between
market value and "equity-plus-knowledge assets" remained almost constant over
that period, smoothly gravitating around two. This supports Strassmann's
(1997, p. 13) position that the market perceives the accumulation of knowledge
assets, which is reflected in the high correlation between share prices of
organizations and their knowledge assets, even though the knowledge assets are
not shown on a company's balance sheet:
The sustained stability of the market-to-capital ratio, which accounts for
the steady rise in the knowledge capital of Abbott Laboratories confirms that
the stock market will recognize the accumulation of knowledge as an asset even
though the accountants do not. The stock market will also reward the
accumulators of knowledge capital because investors recognize that the worth
of a corporation is largely in its management, not its physical of financial
When we move from a macroeconomic to a microeconomic perspective and look
at the business processes of a firm, the trend toward valuing knowledge seems
to be similar to the one just described. Knowledge allows for the prediction
of process-related outcomes, from the more general prediction of acceptance of
a new product by a group of customers to much more specific predictions, such
as slight manual corrections needed on a computer board surface after it goes
through an automatic drill. Correlational knowledge enables process-control
workstation operators at a chemical plant to link a sudden rise of an acidity
gauge to an incorrect setting of the flow through a pipe valve. This enables
the operators to take the appropriate measures to bring the acidity level down
The workers who hold bodies of expert knowledge are rewarded according to
their ability to perform process activities in an efficient and effective way.
This is typically done through linking different types of information, which
can be done through formal educational or personal experience (i.e., the
building of mental knowledge bases), and generating information about the
future based on information about the past or present (i. e., predicting the
future). Organizational wealth is closely linked to the ability to build and
use technological artifacts to control future states of the (economic,
physical) environment in which organizations operate. However, this control is
impossible without the related ability to predict the future, which, in turn,
relies heavily on predictive knowledge.
Organizational knowledge is believed to be the single most important factor
that ultimately defines the ability of a company to survive and thrive in a
competitive environment (Davidow and Malone, 1992; Drucker, 1995). This
knowledge is probably stored mostly in the brains of the workers of an
organization, although it may also be stored in computer systems and databases
(Alster, 1997; Strassmann, 1996; 1997) and other archival records (e.g.,
Whatever form it takes, knowledge is a commodity; and, as such, it can be
bought and sold, which makes its value fluctuate according to the laws that
regulate supply and demand. Abundant knowledge, which can be represented by a
large number of available professionals with the same type of expertise,
becomes cheap when supply surpasses demand, which is typically reflected in a
decrease in the salaries of some groups of professionals. On the other hand, a
situation in which some types of highly specialized knowledge are in short
supply, while demand grows sharply in a short period of time, can lead the
knowledge holders to be caught by surprise when faced with unusually high bids
by employers. For example, Web Java programmers were being offered salaries of
up to $170,000 early in 1996, even though the demand for their new expertise
was virtually nil until 1995. This was the year Java was first released by Sun
Microsystems and 2 years after the University of Illinois began the
distribution of its World Wide Web browser