We can now examine in more general terms the kind of expectations model that is suggested by the studies reported above and by our present under-standing of human organization problem solving. As before, we consider the organization to be a coalition having a series of more or less independent goals imperfectly rationalized in terms of more general goals. From time to time this coalition (or parts of it) makes decisions that involve organizational resources of one sort or another. These decisions depend on certain information and expectations formed within the organization.
From the point of view of such a coalition, the classic theory of expectations is somewhat awkward. In fact, we would expect that an organization that attempted to obtain the information required in such a theory would ordinarily incur heavy costs in internal conflict. An organizational coalition does not require either consistency or completeness in information; in fact, as we have already seen in our case studies, consistency or completeness would, at times, create problems in finding feasible solutions.
At the same time, we do not expect to find — and do not find — anything like a constant level of search. Rather, there are search procedures called into play on various cues such that for any given situation there is a standard search response. In the extreme case, we may want to say simply that either there is search or there is not; there is search when existing decisions are perceived as inadequate. More elaborately, we will want to allow several equivalence classes representing different intensities and types of search. For example, we will want to specify a hierarchy of search activities. If one fails, we proceed to the next.
More generally, the analysis of these studies suggests some problems with both the neoclassical conception of organizational use of expectations and some recent suggestions for revision. On balance, however, it suggests that many of the criticisms of the conventional theory of decision making are warranted, at least in part.
Resource allocation within the firm reflects only gross comparisons of marginal advantages of alternatives. All the decisions studied in actual organizations were made within budgetary constraints and to that extent reflected any marginal calculations that entered into the formation of a general budget. When the rising estimates of costs for the crane controllers (Decision 1) and the decline of business created an internal problem of scarce resources, there were some attempts to compare the advantages of the safety devices with alternative investments. These attempts, however, focused on such considerations as prior commitment rather than marginal return.
In the other cases, there were distinct conceptions of “appropriate” costs or net return. Undoubtedly these were related in a relatively unsystematic way to the comparable statistics (e.g., “payoff period”) on the other acceptable alternatives. If we ignore for the moment the problem of bias in estimates, the studies indicate that rules of thumb for evaluating alternatives provide some constraints on resource allocation even though the allocation is substantially decentralized and there is no conscious comparison of specific alternative investments. Thus, a theory that predicts grotesquely large deviations from a return on investment norm is probably not accurate.
On the other hand, it seems clear that the constraints do not guarantee very close adjustment, especially where business conditions permit organizational slack. Any alternative that satisfies the constraints and secures suitably powerful support within the organization is likely to be adopted. This means that decision making is likely to reflect a response to local problems of apparent pressing need as much as it will reflect continuing planning on the part of the organization.
In a rough sense we can say that the first two case decisions considered here arose primarily as responses to “crisis” situations, the last two as the results of planning. In the computer decisions the organization had been alerted to the potential utility of electronic data-processing and had actually instituted procedures for continuing attention to possible applications. In the other cases the organization was stimulated to search for solutions to conspicuously unsatisfactory conditions. In every case, once an alternative was evoked, it was accepted if it satisfied the general cost and return constraints and enjoyed the support of key people in management. This support in turn came about through a rather complex mixture of personal, suborganizational, and general organizational goals. In Decision 2, the support came for mutually contradictory reasons from two or three different parts of the organization. In Decision 1 the support from top management came for reasons that were not directly relevant to the events that had triggered the search. In Decision 4 the support from top management came in considerable part from collateral expectations about the action.
Search activity is not viewed as simply another use of internal resources. In general these studies suggest that there are several stages to motivated search activity on the part of an organization. If a problem area is recognized, there is ordinarily a search for possible alternatives. At this stage only rough expectational data are used to screen obviously inappropriate actions. In each case considered this early scanning generated only a few suitable possibilities, which were then considered in greater detail. In most cases a rather firm commitment to an action was taken before the search for information proceeded very far, but the search became more and more intensive as the decision approached implementation. This was especially obvious in the case of the crane controls.
One major reason why this seems to be true is that organizational “search” consists in large part of evoking from various parts of the organization considerations that are important to the individual subunits; the relevance of such considerations, and the impetus to insist on them, are not manifest until the implications of the decision are made specific through implementation. An obvious corollary of such a conception of the search process is the proposition that search will be much more intensive where organizational slack is small than where it is large. Where there are enough excess resources in the organization, the interdependence of allocation decisions is uncertain; search consequently becomes relatively routine.
At the same time a conspicuous factor in these cases is frequently ignored in search theory. Whether in its classical form or in the level of aspiration form, the theory of search is basically a prospecting theory. It assumes that the objects of search are passive elements distributed in some fashion throughout the environment. Alternatives and information about them are obtained as a result of deliberate activities directed toward that end. Not all information comes to an organization in this way, however. Many of the events in these studies suggest a mating theory of search. Not only are organizations looking for alternatives; alternatives are also looking for organizations. In the computer decisions the intensity of search activity by the organization would scarcely have generated as much information as it did if the manufacturers of electronic data-processing equipment and the consulting firms had not been pursuing as well as being pursued. In Decision 1, too, many efforts had presumably been made by producers of magnetic controllers to sell management on a change. In fact, the timing of the major spurts of activity on those decisions was as much a function of the pressure from such outside groups as it was from internal factors.
Computations of anticipated consequences used by the organizations seem to be quite simple. Although there is no particularly strong evidence for Shackle’s specific concept of what computations are made, at most only a half-dozen criteria were used explicitly in making the decisions. There appear to be two main reasons for the simplicity. First, in one form or another the major initial question asked about a proposed action was not how it compared with other alternatives but whether it was feasible. In the decisions discussed here there were two varieties of feasibility. The first was a budgetary constraint: Is money available for the project? The second was an improvement criterion: Is the project clearly better than existing procedures? In one form or another these questions were extremely important in all of the decisions discussed here. In some cases they were rather hard to answer, but they were almost always considerably easier than the question required by the classical theory of expectations: Does the expected net return on this investment equal or exceed the expected return on all alternative investments?
The second apparent reason for simplicity in establishing decision criteria was the awkwardness of developing a single dimension on which all relevant considerations could be measured. In each of the decisions described above, costs in dollars were factors; so were dollar savings. But so, too, were such considerations as speed and accuracy of work, safety of personnel, distance from railroad transportation, quality of performance, and reputation of company. Unless one is prepared to make explicit the dollar value of such diverse factors — and none of these firms did do so to any great extent — they must be treated substantially as independent constraints. Detailed expectations on these dimensions seemed substantially irrelevant to efforts to estimate costs because the organization had no way of using such information.
Expectations are by no means independent of such things as hopes, wishes, and the internal bargaining needs of subunits in the organization. Information about the consequences of specific courses of action in a business organization is frequently hard to obtain and of uncertain reliability. As a result, both conscious and unconscious bias in expectations is introduced. In each of the cases studied there is some suggestion of unconscious or semiconscious adjustment of perceptions to hopes. The initial estimates of cost for the crane controllers appear to have been fairly optimistic. The expectations about the consequences of moving the home specialties department seem to have been substantially a function of subunit goals. The evaluation of consulting firms seems to have shifted before detailed expectations were formed; subsequently, the expectations supported the evaluation. Expectations about net return from alternative data-processing systems apparently were influenced by some feelings of a priori preferences.
In addition, there is some evidence of more conscious manipulation of expectations. The classic statement came from a staff member involved in one of the decisions. He told a group of men outside of the company, “In the final analysis, if anybody brings up an item of cost that we haven’t thought of, we can balance it by making another source of savings tangible.”
It would be a mistake to picture the biases introduced in either of these fashions as exceptionally great. In almost every case there are some feasonably severe reality constraints on bias. But where the decision involves choice between two reasonably equal alternatives, small biases will be critical. Consequently, research on selective perception and recall is of substantial importance to an empirical theory of business decision making.
Communication in a complex organization includes considerable biasing, but also considerable bias correction. The communication system is not passive, nor is it viewed as passive. Information on critical variables in the system is seen as an important element in the decision process. In both of the experiments on communication, organization members modified their communicated judgments in the light of their picture of the decision consequences of various information.
Despite the bias in communication introduced in this manner, however, the system does not become hopelessly confused. Most biases are recognized by other parts of the organization. Whether the recognition is a logical inference drawn from the goals in the organization or an experienced inference drawn from past cases, long-run biases are detected. Once detected, the bias is vitiated by a correction factor. In addition, organizations seem to protect themselves from the worst effects of bias by focusing on easily verified data in lieu of uncertain estimates and by using easily checked feedback information instead of more remote anticipations.
Source: Skyttner Lars (2006), General Systems Theory: Problems, Perspectives, Practice, Wspc, 2nd Edition.