Empirical study of decision-making quickly revealed that three basic components of the process were absent from the classical theory. One omission is the process of setting the agenda that determines what deci-sions will be made at what particular times. The second is the process of obtaining or constructing a representation for the problem selected for attention. The third is the set of processes that generate the alternative actions among which the decision-makers choose. These processes call for more elaboration than they receive in Chapter V.
1. Setting an Agenda
In the classical theory, it is supposed that the same set of decisions is made at each point in time. In that theory there is no such thing as an agenda, for there is no need to choose which specific decision problems will be dealt with. In the real world, the available attention must be directed to those matters on which timely action is required instead of those about which there is no urgency; there must be processes for setting and revising the agenda.
Simple Procedures for Setting Agendas. If two or more needs express them- selves at the same time, organisms and organizations must decide which to put first on the agenda. These- priorities are usually settled by simple rules: attend first to the need whose inventory of satisfiers is more nearly exhausted. Agendas are set very much in the manner of the familiar two- bin inventory systems of industrial practice. For each need or want there is an “order point” and an “order quantity.” At some level of deprivation, sig- nals sent to decision centers secure attention to the want unless more urgent signals are present. If the matter is not attended to immediately, the signals become gradually more insistent until the want gains first priority.
This system for fixing the agenda requires nothing like a comprehensive utility function. The urgency of needs is compared only to set search priorities. All that is required is a simple mechanism that will signal urgency and gradually increase the intensity of its signal. Nothing needs inventories are completely exhausted. With some slack available, searches can be interrupted in the face of more urgent demands.
A mere increase in the number of issues for decision does not com- plicate the agenda-setting task, provided that attention to all of them is not essential for survival. Those that are not urgent enough simply never get on the active agenda. (Most of us are familiar with this phenomenon in our personal lives.) Most potential agenda items are either problems or opportunities. Problems are items that, if not attended to, will cause trouble. Opportunities are items that, if attended to, may increase profit or probability of surviving.
Nor is there a definite list of opportunities, or even problems, among which the priorities are set. Neither problems nor opportunities can be considered for the agenda unless they are noticed, and except for those that attract attention by means of an internal signaling system, they must be picked out from a complex external sensory environment. Until they are noticed, opportunities are not opportunities. In the world in which we actually live, at any given time we notice only a tiny fraction of the opportunities that are objectively present, and only a small part of the problems. A major initial step—and by no means an assured one—in technological or social invention is to extract opportunities and problems from the confusion of the environment—to attend to the right cues.
How Opportunities Are Noticed. Today, we have the beginnings of a theory of how opportunities ( or problems) are noticed. The greatest progress has been made in the domain of scientific discovery.29 One of the mechanisms that focuses human attention on important problems is surprise. Alexander Fleming noticed a Petri dish in his laboratory in which the bacteria were disintegrating. He was surprised—there was no obvious reason why the bacteria should be dying. On the edge of the dish, near where the lysis was occurring, was a mold of the genus Penicillium.
What are the conditions for such a surprise? We are surprised when we are knowledgeable about a situation and something unusual (contrary to our knowledge) occurs. Fleming was knowledgeable about bacteria and molds, and nothing in his knowledge led him to expect that bacteria would die in the presence of a mold. Surprise put the problem (or oppor- tunity) of explaining why the bacteria were dying on Fleming’s research agenda; it would not have been noticed by anyone who lacked his knowl- edge. A great many opportunities, including many of the first order of magnitude, secure their place on the agenda through informed surprise.
We can generalize from the surprise mechanism to a more general theory of what it is that focuses human attention on specific parts of the environment. In the contemporary world all of us are surrounded by, even drowned in, a sea of information, only an infinitesimal part of which can be attended to. Although we may wish to have certain kinds of information that are not available (e.g., reliable forecasts), the critical scarce factor in decision- making is not information but attention. What we attend to, by plan or by chance, is a major determinant of our decisions.
Given the general scarcity of attention, people and organizations can enhance the quality of their decision-making by searching systematically, but selectively, among potential information sources to find those that deserve most careful attention, and that might provide items for the agenda. This is a major function of so-called “intelligence” units in organizations, and also of research and development units, and even planning units.
For example, a company laboratory is seldom the major source of basic discoveries from which new products can be developed. More often, the laboratory serves as an intelligence link to the community of academic and other science from which ideas may be drawn. Its task is to observe and communicate with that community, and to notice and develop further the opportunities that are presented by it. Of course, the experimental laboratory also has its window on the natural world, but that is a rather narrow window unless supplemented by close interaction with the scientific community.
A common responsibility of planning units, not always explicitly rec- ognized in the definition of their function, is early recognition of problems. One mechanism for problem recognition is to build computational models of the system of interest and use them to make predictions. Selective surveillance of information available in the environment may provide an even more reliable early warning system than prediction.
Perhaps I have said enough to demonstrate that a theory of agenda formation—which is, in turn, a theory of attention focusing—is an essential part of a theory of rational decision. We can find ideas on this topic useful for decision-making in the literature of artificial intelligence and cognitive science—for example, recent research on the processes of scientific discovery.
2. Representing the Problem
The commentary on Chapter II noted that an organization’s structure is itself a representation of the task the organization was designed to deal with. Representation also has significance at the level of decision-mak-
Perhaps even less is known today about the mechanisms of problem formulation than about agenda-setting processes. Of course, if the item placed on the agenda by the attention-directing mechanisms is of a familiar kind, standard procedures will usually be available for casting it in a solvable form. If we can formulate a problem as an equation, then we know how to solve it.
Or, to return to items placed on the agenda by surprise, scientists have a rather standard procedure for exploiting surprises. In case of surprise, they first try to characterize the scope of the surprising phenomenon. If bacteria are dying in the presence of a mold, what kinds of bacteria are affected? (Fleming found that many kinds were.) What kinds of mold? (Evidently only the mold Pénicillium.) And when the scope of the phenomenon has been defined, try to find its mechanism. (Can we extract from Pénicillium, by crushing, treating with alcohol, heating, crystallizing, etc., a substance that retains, or even enhances, its effect upon bacteria? If we find such a substance, can we purify it and characterize it chemically? A whole sequence of experiments, first by Fleming, then by Howard Florey and Ernst Chain, achieved just this.)
Some problems are very hard as the world presents them, but very easy when they are reformulated properly. The Mutilated Checkerboard problem is a celebrated example. Consider a checkerboard (eight-by- eight) and 32 dominoes, each of which covers exactly two squares of the board. Clearly, we can entirely cover the checkerboard with the dominoes. Now suppose that two squares are cut out of the checkerboard— the upper left corner and the lower right corner. Can we cover the remaining 62 squares with 31 dominoes? We cannot, but the answer is not obvious. None of us would have the patience to demonstrate the impossibility by trying all possible coverings; we must find some other way. Let us abstract the problem, considering just the number of dominoes, the number of black squares, and the number of red squares. Each domino will cover exactly one black and one red square. But the two squares we removed are of the same color (they are at opposites ends of a diagonal). Hence there will now be two fewer squares of one color than of the other (let’s say 30 black and 32 red). But dominoes can cover only the same number of black and red squares, hence a covering is impossible.
Problem representations, like the problems themselves, are not presented to us automatically. They are either retrieved from memory, when we recognize a situation as being of a familiar kind, or discovered through selective search. Formulating a problem is itself a problem-solving task.
Eastern competition. The problem is on the agenda, but finding an appropriate problem representation is difficult, and has not yet been fully achieved. Is the problem one of quality control, of manufacturing efficiency, of managerial style, of worker motivation, of wage levels, of exchange rates, of foreign trade regulations, of investment incentives? The list is endless; and different representations of the problem will produce different proposals for solution.
It is apparent that developing a veridical theory of problem representation must stand high on the agenda of decision-making research.
3. Discovering and Choosing Alternatives
One of the striking features of the theory of the rational economic man is that all of the alternatives among which he chooses are given at the outset. He lives in a static (imaginary) world that presents a fixed repertory of goods, processes, and actions of every sort. This classical view of rationality provides no explanation of where alternative courses of action originate; it simply presents them as a free gift to the decision-makers.
Yet, a very large part of the managerial effort in any organization is devoted to discovering possible alternatives of action. To take some obvious examples, there is search for new products, for new marketing methods, for new manufacturing methods, even for new organization structures. All of this search activity is aimed at enabling the organization to go beyond actions that are already known and understood and to choose novel ones.
Even Chapter V, which includes some discussion of alternative-finding under the heading of “The Planning Process,” gives rather short shrift to the subject of generating alternatives, and we must count this as a serious shortcoming in our treatment of decision.
House-hunting and job-hunting are market activities that normally require extensive search among an ill-bounded set of alternatives. A graduating student, searching for a first job, must not only have procedures for discovering prospective employers, but stop rules for determining when the search should end, and procedures for obtaining relevant information about each employment opportunity. In just the same way, in organizations the alternatives for choice are not usually given but are generated through selective search.
Finding alternatives is sometimes a search of the sort just described for a house or a job. Here the alternatives already exist; they must simply be located. But in many cases, including perhaps the most important, the alternatives for which an organization is seeking do not exist but have to ket or to be sold from the shelf, but are designed specially on contract with a particular customer. And shelf goods, too, initially have to be conceived and designed, a task that becomes central and continual in industries, like clothing or pharmaceuticals, where new products are constantly coming onto the market.
In recent years, research in cognitive science has taught us a great deal about the processes of design.30 In any problem-solving process, we have a goal, or a set of goals, formulated as tests to be applied to prospective solutions. (A solution is something that satisfies these tests of goal satisfaction.) Design calls for a generator that produces prospective solutions. If it cannot simply produce items, one by one, for test and acceptance or rejection, it must synthesize prospective solutions in a series of steps, applying tests of progress along the way to direct the search. The more we know about the problem space in which we are searching (the problem representation), the more information we can extract from that space to direct the search, and the more efficient the exploration will be.
4. Stages in Decision-Making
The division of the decision-making process into such subprocesses as setting the agenda, representing the problem, finding alternatives, and selecting alternatives has sometimes been criticized as describing decision-making falsely as a “linear” process, and thereby rigidifying it.31 Of course there is no implication in anything that we have said that these subprocesses must follow in a set order. Agenda-setting and resetting is a continual process, as is the search for new decision alternatives (e.g., new products), and the selection of alternatives as new occasions for decisions arise. An alternative discovered in one decision process may find its effective application at some much later time and in connection with a quite different decision.
Moreover, each of the subprocesses in decision-making itself poses a problem that may again require agenda setting, finding alternatives, selecting them, and evaluating them. This becomes clear when we automate decisions in computer programs and observe the complex hierarchy of goals and subgoals that emerges in the course of executing them. There is nothing “linear” about the decision process described here, nor any barrier to flexibility as new situations arise and new facts are discovered. All of this is pointed out, if briefly, in Chapter V.
5. Well-structured and Ill-structured Problems
The problem-solving we understand best concerns well-structured problems. Problems are well structured when the goal tests are clear and easily applied, and when there is a well-defined set of generators for synthesizing potential solutions. Problems are ill structured to the extent that they lack these characteristics. Many, if not most, of the problems that confront us in the everyday world are ill structured. An architect designing a house, an engineer designing a bridge or a power-generating station, a chemist seeking a molecule with desired properties and a way of manufacturing it cheaply, a manager judging whether a new factory should be built to meet increasing demand—all of these are solving problems with many ill-defined components.
To the best of our current knowledge, the underlying processes used to solve ill-defined problems are not different from those used to solve well- defined problems. Sometimes it is argued, to the contrary, that solving ill- defined problems involves processes that are “intuitive,” “judgmental,” or even “creative,” and that such processes are fundamentally different from the run- of-the-mill, routine, logical, or analytical processes employed in well- structured problem-solving.
We can refute this argument empirically, because we have strong evi- dence today about the nature of intuitive, judgmental, and creative processes that shows how they are carried out. We know that experts in any domain have stored in their memories a very large number of pieces of knowledge about that domain. Where it has been possible to measure the knowledge, at least crudely, it appears that the expert may have 50,000 or even 200,000 “chunks” (familiar units) of information—but probably not 5,000,000.
This information is held in memory in a particular way: it is associated with an “index”—a network of tests that discriminate among different stimuli. When the expert is confronted with a situation in his or her domain, various features or cues in the situation will attract attention. A chess player, for example, will notice such familiar cues as an “open file,” “doubled pawns,” or a “pinned knight.” Each familiar feature that is noticed gives access to the chunks of information stored in memory that are relevant to that cue. An accountant who sees a low cash balance on the balance sheet will be reminded of what he or she knows about cash flow and liquidity problems.
The ability, often noticed, of the expert to respond “intuitively,” and often very rapidly, with a relatively high degree of accuracy and correctness, is simpiv the product of this stored knowledge and the problem- upon experience and knowledge. There is nothing more mysterious about them than about our recognizing our friend “instantly” when we meet him on the street, and gaining access to all sorts of information we have about that friend. However, as the ideas of intuition, judgment, and creativity are widely believed to be beyond scientific explanation, we will take some pains in the next section of this commentary to say more about what is known about them.
As we shall see, we do not need to postulate two problem-solving styles, the analytic and the intuitive. The power of analysis depends on expert knowledge for its speed and effectiveness. Without knowledge that becomes available by recognition, only tiny, slow, painful steps can be taken in reasoning. We may see relative differences among experts in their reliance on analysis as against recognition (intuition), but we may expect to find large components of both, closely intermingled, in virtually all expert behavior.
Source: Simon Herbert A. (1997), Administrative Behavior, Free Press; Subsequent edition.