During the 1960s, when the first generation of useful all-purpose computers reached the market, a kind of decision support called electronic data processing, EDP, became available. It was designed to do the task of implementing decisions which had already been made. Very large quantities of data were processed, each in accordance with a well-defined and programmed series of choices and conditional branches. In a loose sense this processing can be considered to have comprised decisions, although the ‘decisions’ were nothing more than mechanical recognition of patterns without the aid of human judgement. Its designers were typical computer scientists and it was applied in transaction systems, among other uses.
Later, the concepts of Management Information Systems, MIS, emerged. These systems were the result of a cooperation between computer science and the area of management. The underlying need grew out of an increased internationalization and competition among organizations in a more complex world. Visions of computers revolutionizing the business sector with online, real time systems supporting rational, quantitative decisions by large amount of data were common. In reality the MIS has become a tool for routine middle management decisions by supporting simple bookkeeping, updating inventory and calculating cost over-time, and so on.
As a predecessor to MIS, the Decision Support System, DSS, arrived in the 1970s as an instrument primarily for modelling. Also, this system was intended to be a tool for decision makers, but became a little too complicated for the average executive. DSS systems were generally geared for special, limited processes of decisions involving multidimensional models, ad hoc analyzing, reporting, and consolidation. Today, these systems are used mainly by controllers and analysts for calculating, budgeting, simulation and aggregating.
During the 1980s, two generations of Management Support Systems, MSS, were introduced. The first generation comprised mainframe computer systems with links to personal computers. A private database was included in the system, which had a menu-driven, character-based user-interface. The system-support was manual, usually administered by a special computer department. The software consisted of different software packages for different functions.
At the end of the decade a change of generation took place among management computer systems and a second generation MSS emerged with the client/ server technology. No particular databases were used and communication took place with databases belonging to other, external, agencies. The user-interface became a graphics screen used for data-driven models, consolidations and reports. The software packages were integrated for all functions.
At the beginning of the 1990s the demands on computer systems grew rapidly. From the executive’s viewpoint the data provided was produced too slowly and was not good material for calculations. With the rapid technical development that led to powerful workstations, relational databases and improved data communication, new demands arose. These demands were met by the concept of the executive information system, EIS, which emerged in response to the demands. (Some computer scientists use the acronym ESS, executive support system which is synonymous with EIS.)
The idea was that EIS should be a support for top-level executives in their daily work involving decision-making, planning, and controlling. Information should therefore be presented, processed, and handled in a very simple manner appropriate for the layman. The main point of the system should be flexibility and user-friendliness.
A closer look at EIS shows that the system is customized for a particular decison maker and it is used directly by that decision maker without an intermediary. The system is extremely user-friendly, requires relatively little learning time and gives the user-support for all kinds of problem definitions. It is designed to process large amounts of data from both internal and external sources and gives the user the possibility to predict and simulate various courses of action.
EIS may be described as a computer supported presentation/ analysis tool which collects data from other systems or from its own database. It facilitates drill-down, that is, the selection of information and navigation from an aggregated to a more detailed level, in a well- structured way. Requested data is processed by the user according to his wishes. Its main functions lie in analysis and modelling where information from different sources is combined. Information can also be distributed by the system to employees in other departments by electronic mail. The presentation of information is achieved via high-resolution graphics on screens and by tables and text.
Today, decision support is also found in expert systems, ES, which can adapt their own rules in a manner predetermined by another set of rules (see p. 332). By the use of artificial intelligence and by imitating processes that human experts unconsciously perform, the system can serve as an advanced problem solver and decision maker. Note, however, that the problem to which an ES is applied must be complicated enough to require an expert. It can also be worth to note what the well-known physicist Niels Bohr said about experts.
“An expert is a person who has made all the mistakes that can be made in a very narrow field”.
AI systems, which must be considered a tool for more loosely structured problems, can be characterized as inductive, synthetic and trial-and-error based. When there are no well-defined problems or no consensus exists regarding the right solution, an ES makes no sense. ES are typically used where expertise is unavailable, scarce or too expensive in order to manage inconsistent, incomplete or uncertain information. Another typical area of use is where time and pressure constraints are involved and a need exists to secure knowledge before an expert person quits a company. Time-consuming problems where several days are required for a solution are not suitable for ES systems. In principle, the expert system can deliver its own solution at decision time and do the whole job itself, although a responsible decision maker is unlikely to give up his right to decide.
Table 9.5 shows how the various kinds of computerized decision support systems have been used at different management levels.
From Table 9.5, it is clear that DSS is especially adapted to support decisions at the tactical level and in some cases at the strategic level. Observe that an ES is not suitable for solving unstructured problems; this type of problem occurs frequently at the strategic level and requires human judgement and skill at the moment of decision.
Decision support systems are either data-oriented or model- oriented. Data-oriented systems could be, for example, online budgeting systems while model-oriented systems can be exemplified by an accounting system which calculates the consequences of a particular action. When designing systems for structured tasks (MIS) the approach is mainly analytic, planned and deductive. The creation of systems for more loosely structured problems (AI systems) can be characterized as inductive, synthetic and trial-and-error based.
The general structure of a system is shown Figure 9.2.
The user-interface for the dialogue management unit consists of a workstation with a set of programs which manage the display screen. It obtains input from and sends output to the user and translates the user’s requests into commands for the other two units. The model management unit contains models of the business activity. Examples are spreadsheets, financial models and process simulation models.
Figure 9.2 General structure of a decision support system.
This unit also creates, modifies and invokes the models. The data management unit maintains the internal database and interfaces other sources of data from external databases.
Something typical for an ES is that the control structure is separated from its data-content. The reason to separate the general reasoning mechanism from all the stored facts and heuristics is that the system then can be incrementally built-up and refined on a continuous basis. Such a structure also facilitates a testing of the whole system.
An important feature of an ES is that it is able to comment its reasoning like a human expert by an explanation facility. It is thereby possible to ask how certain facts were considered and how the system delivered the specific conclusion. Involved in this process is the possibility to handle uncertainty which is allowed to be entered in the system when consulted.
An ES delivers its solutions according to three main strategies. When the starting point emanates from concept formulation or design, a reasoning procedure called forward chaining is used. In lack of particular goals, the computer here processes and analyses all accessible data and gives a solution. This can sometimes be very random or overloaded as everything possible could be presented. Backward chaining is a kind of goal-directed reasoning starting with a hypothesis. The search is done backwards along a certain track to verify the hypothesis. This strategy is often used in planning; a drawback which sometimes occur is, however, a combinatorial explosion of possibilities. Combined chaining is mainly used when complex problems with pronounced uncertainties exists. An example is oil drilling prospecting.
Source: Skyttner Lars (2006), General Systems Theory: Problems, Perspectives, Practice, Wspc, 2nd Edition.