In a complex management environment, the events to be controlled cannot be comprehended within a reasonable time. The different aspects of the events, however, may be grouped into modules, which can be controlled as one unit.
This type of control is influenced by the dependence of modules. This dependence may result where the controlling of one aspect limits the choice in another aspect. If these limitations exist within one module, the control is not impaired; if they exist from module to module, the modules are dependent ('bound foreign aspects' are not controlled by us, but they limit the choice in the control of our module).
Decision making for a module demands data, both for the 'indigenous' aspects of the module and for its bound foreign aspects. The data take the following form: 'reports' describe the situation to depart from, 'orientations' the decisions from the modules that limit our choice, 'extrapolations' the most probable outcome of uncontrolled aspects. These three types of data are combined and evaluated to form 'predictions' which describe the course of events as accurately as possible.
For every module there also is a 'statute' containing the rules (prescribed or discretionary) that govern the actions of the controlling unit of the module. The statute and predictions are combined to yield a 'decision' that is published as a plan insofar as it implies limitations elsewhere. After some delay, necessary for the dependent modules to react to this plan (for them an orientation) the decision is executed or implemented (to be executed by others).
The 'term' of a module is the average time needed to initiate the execution of a decision with respect to the aspects in that module. A close relationship must exist between the term of a module and the time needed to change the initial situation to the desired situation. This relationship is used to determine the allowable term of a module and hence, depending also on the kind of aspects, its size.
The formation of modules leads to a management structure that is subject to a number of efficiency criteria: viz. economy, completeness, consistency and 'power' (the number of aspects that are controlled within a given time).
The enhancement of the 'power' of a management unit ought to be considered as the primary objective of measures to change the structure or to accellerate the data processing.
Structural measures consist of lodging new aspects in modules or of joining and splitting existing modules into new ones. It appears necessary to continually reorganise a management structure.
A management structure is not homogeneous; there are 'lower' and 'higher' modules. The latter comprise those aspects which cause numerous limitations and are therefore considered important.
Important aspects usually are concurrent with a long term of the module: this is not always the case. The relations between management units are determined by the relative superiority of their decisions: superior meaning 'close to reality, workable' rather than 'close to an ideal'. Units that frequently publish superior decisions become dominant in the structure; the control of higher modules must be realised by dominant units.
The prescribed part of the statute of a module originates with a number of dominant modules; not with one 'boss'. In this concept of a management structure there is no place for notions like 'span of control', 'vertical hierarchy', 'staff and line functions' etc. In our idea, a structure appears as a dynamic, pyramidical form in which a number of modules exist with multiple connections of different kinds for the exchange of information.
The term of a module is partly determined by the information technique employed in and between the management units. Fast processing of data leads to bigger modules, more power, a better approximation of completeness and perhaps to more economic control. These results can only be reached if the better information technique is supplemented by the appropriate structural changes.
Data processing is an ubiquitous activity: it appears useful to define some of its concepts in detail. First we introduce an information system as being a set of data with a corresponding separation function to divide sets of data into three subsets: useful within the system; useful outside the system, useless. We then recognize 'observations' (data to be appended to the system) 'messages' (data to be released by the system) and 'experience' (data retained by the system at a given moment). Observations contain information, if the experience is changed as a result of the processing of that observation. Within the set 'experience' we recognize the subset 'procedures'; it is the description of all functions that can be applied in the system. Part of the procedures - the construction - cannot be changed without altering the system; the remaining part of the procedures - the instructions - can be modified within the system.
Some procedures are described exactly; they are programs. Other procedures are described insufficiently, or not at all; some are even indescribable as yet. Computers can execute all data processing, including modification of instructions, provided the procedures are nothing but programs.
The abstract formulation of the concept information system leads to the mathematical descriptions of system, subsystem, union of systems, and intersection of systems. An algebra of systems can now be developed for stationary systems (no modification).
An interesting kind of information system is the control system, that contains in its experience a realised model of a module of aspects. The observations of a control system concern the indigenous and bound foreign aspects of the module, the messages give rise to the fixing of the indigenous aspects. The definition of module (assuming the simultaneous control of all aspects) implies that control systems cannot have a union and cannot be split. Subsystems of control systems, however, can be integrated into a union of systems.
The information technique as it exists in a management structure can be improved in several ways. Qualitative improvement can be made' by the application of refined models, by a decrease in the delays that render predictions useless and by an increase in the consistency of reporting. Quantitative improvement can be made by an increase in data processing speed. Fast information techniques can result from three measures: standardization, integration, mechanization; these measures are sometimes interchangeable and, moreover, partly interdependent.
Standardization is the effort to use identical components in different systems; integration is the development of systems that comprise a number of (sub)systems; mechanization is our word for the application of computers in information systems. Each of these measures has advantages and disadvantages, not necessarily all systems must be standardized, integrated and mechanised. The concept 'total system' is not useful.
Integration appears efficient where systems serve dependent modules and have common components (files or procedures). The time needed to develop an integrated system, however, can be too long with a view to the rapidly changing demands. (In the next decade this is the case.)
Computers offer the opportunity to design systems with intricate procedures and yet a short response time. Through the use of remote and multiple input/ output equipment and a central memory the consistency of the data can be insured. An integrated, mechanised information system for a management structure (MIB) can supply the reports, the orientations, the extrapolations and the analyses of tentative decisions for a number of control systems. The human element in such a system demands that the computer element responds immediately to tentative decisions. A constraint of this nature suggest a random access memory organisation: one possibility is to employ the chain addressing method.
The specifications of the computer for a MIB are such that we may expect these systems in the near future.