The construct of bounded rationality provides a major link between IT and organizational design. Bounded rationality at the individual level refers to neurophysiological limitations to the information processing capacities (memory, computation and communication) of an individual. It is demonstrated in limits on the complexity and size of problems that can be solved by humans. IT can directly affect the computational and communication abilities of a decision-maker, thus shifting the limits of rationality. Bounded rationality has been a key concept in the development of organizational theory. There are human limitations on information processing interact with environmental factors such as environmental complexity and uncertainty to give rise to organizational problems. From this school’s perspective, organizations are designed to marshal sufficient information processing and communications capabilities to be able to manage the complexity and uncertainty in the environment.
Simon emphasized the limitations on human computational capacity, and concluded that “bounded rationality” is the most appropriate way to describe a scientific behavioral grounding of economics. Individual agents are, in his view, to be thought of as “satisficing,” that is, doing the best they can to move in the direction of their goals [Simon 1955, 1979 are representative examples of his approach. For a contemporary review article, see Conlisk 1996.]. The bounded rationality view is in some general sense undeniably true, but its very generality has prevented it from displacing maximization as the dominant mode of theorizing in economics. Satisficing is too inclusive as a characterization of behavior. Bounded rationality per se rules out hardly any behavior---seemingly irrational actions can be attributed to the unavailability or costliness of information, or to ubiquitous (but hardly ever measurable) “transactions costs.”
It is necessary is to be more specific about the limits to “rationality.” The growing literature on agent-based models offers one approach to this specification problem. In agent-based theory, the economic entities operate according to well-defined rules of behavior and on the basis of information available to them. No overreaching maximizing principal is invoked, although “optimal” patterns may emerge through selection and adaptation [Kauffman 1989]. Emphasis is on the calculations the agents actually are capable of performing, given their processing capacities and the data available to them. We propose a model of the adoption of innovations by firms that is in this tradition, and show that it provides a plausible explanation of some of the phenomena that are difficult to reconcile with the profit maximization story.
In the more recent transaction cost view of organizations, Williamson asserts that constraints on human information processing are a major reason for the very existence of organizations. An alternative to organizations is to have economic agents acting independently and contracting to sell their services to one another in a marketplace. With unbounded rationality, every participant could counteract the effect of other participants’ deceptive, self-interested behavior. In a world of bounded rationality, however, such opportunistic behavior in small marketplaces creates inefficiencies in the form of excessive contracting and transaction costs. To avoid these costs, individuals form organizations where interests are pooled. The transaction cost view can provide a new perspective on the role of information systems in organizations.
According to the transaction cost schools of organizational theory, bounded rationality plays a pivotal role. Therefore, to explain the role of IT in improving internal strategy, we must characterize systems in terms that are relevant to bounded rationality. For example, instruction processing speed and range and depth of system functions are different aspects of possessing capacity but they are not equally relevant in a bounded rationality view of organizations. The former measures only technical capacity, whereas the latter can be directly linked to reducing human limitations in organizations.
Organizationally relevant measures of system:
Functional components: Storage, processing & communications
Performance characteristics: Capacity, quality and unit cost
The major premise of this discussion is that IT affects the efficiency and effectiveness of the organization primarily by reducing the bounded rationality of individual and group decision-making.
The second fantasy is that organization is information. This is really an extension of the first fantasy to the organizational level and claiming that the organization as a whole is structured data. Galbraith’s (1977) Organization Design is the best-known attempt at making full use of this type of imagery and demonstrating how the organization can be managed as one large information system. Although very elegant in it's reasoning it is totally abstract in terms of how real life organizations can be managed.
New organizational structures are created to close the information gap between requisite information processing needs of a firm and available information capacity. Bounded rationality is controversial and obscure concept for the economists. While it is easy to see what bounded rationality denies- namely the substantive rationality of full optimization – it is not clear exactly what it substitutes for it. Simon’s (1947) alternative is procedural rationality, which is akin to the idea of using optimal procedures in a world of costly information. The assumption of bounded rationality has been replaced by an assumption that information is costly to collect and to communicate. This new assumption is both more powerful and more precise. It is more precise in the sense that it avoids ambiguity over exactly what deviations from substantive rationality are liable to occur.
The approach based on information costs integrates the analysis of the employment contract with the analysis of the organizational structure of the firm and hence powerful too. Transaction-cost economics traces the origins of information costs to bounded rationality, which is omnipresent, and seeks to relate the incidence of information costs to characteristics of transactions and to the informational demands placed on decision makers under alternative organizational forms. Thus, modern transaction-cost economics regards information and other transaction costs as something to be explained rather than assumed, a difference that accounts in large measure for why transaction-cost economics, unlike information and game theories, has generated a substantial body of empirical work using real-world data.
Conventional analyses are content to treat information, bargaining and contracting costs as parameters; transaction-cost economics regards these as consequences of human cognitive limitations and behavior. It is true, for instance, that the complications introduced by bounded rationality can be captured in models of utility maximization constrained by imperfect information.
In economic theory, the firm is thought to be a locus of information, as embodied in the production possibility sets. A firm then has an information base but it is typically distributed; not everyone in the firm has every piece, and transmission, even within the firm, is costly. As a result we expect to see specialization of function to economize on the transmission of the information. Specialization leads to the need of internal coordination, but the matter is made more acute and more difficult to analyze when there are differences of information within the firm.
In the literature on "bounded rationality," Van Zandt (1997) provides a model in which firms will optimally decentralize decision-making as the flow rate of new information increases relative to the capabilities of the central information processors.
Simon’s conception of bounded rationality as the basis of the processes of organizational decision, which allows for information technologies to play a role in lifting the constraints of rationality.
The mathematical model that is used to represent this phenomenon is that of the bounded rationality constraint, which states that if the workload rate exceeds some value, rapid degradation of performance occurs. Decision makers normally operate at a level where the bounded rationality constraint is not in effect. The bounded rationality constraint cannot be determined analytically but can be established experimentally.
Since the bounded rationality constraint cannot be determined analytically, an adjustable workload limit is used in the model. The workload limit places an upper bound on the amount of tasks a decision maker can accept in one processing interval. If the decision maker receives a set of inputs that exceeds the workload limit, the decision maker’s status changes to “overloaded”. If new functions are sent to the decision maker while he is in the overloaded status, they are delayed one processing unit, representing the degradation in performance associated with overloaded decision makers. When the decision maker’s status returns to “under loaded”, functions are once again completed in a single processing unit. The local or decision maker monitor provides the mechanism to monitor the workload of the individual decision makers. The output of the monitor, the decision makers’ workload status, is used in the post stage of the decision maker model to determine if a backup decision maker is necessary when the organization is operating under local adaptation, and to indicate when functions should be delayed due to overloaded decision makers.
Given that human decision makers have to operate within conditions of “bounded rationality”, the decision-making processes have to be clearly identified, so that for each decision all the variables are accounted for and the information that has to be processed by the decision maker can be reduced to a minimum. In order to achieve this, Simon developed a model of organizational decision making based on the inner workings of a computer, whereby human beings act as information processing systems which extract meaning structures from information inputs and store such structures as knowledge for later use in decision making. Such meaning structures, however, are seen as static and disembodied in the sense that they are divorced from the stream of organizational actions that produce and reproduce meaning (Tenkasi and Boland, 1993).
In an organization information system is an instrument of three uses: decision making, monitoring and intermediate data processing. Decision making and monitoring are terminal uses of the system; intermediate data processing facilitates a terminal use. The data processing in an organization gives the organization the ability to adapt to the specific decision problems and monitoring needs of the terminal users, hopefully leading to cost savings and /or better decisions.